Land mattered. Labour mattered. Logistics, taxes, suppliers, customers, planning permission all mattered.

Electricity? You connected to the grid.

That assumption is starting to break.

A company can secure the site, approve the capital, line up customers and still discover that its factory, data centre, logistics depot or charging hub cannot operate on the timetable it planned because sufficient electrical capacity is not available.

And that takes us somewhere much bigger than energy policy.

If access to power determines where investment can happen, grid policy starts to become industrial policy.

The scale of the emerging constraint is extraordinary. The International Energy Agency’s Electricity 2026 report estimates that more than 2,500 GW of renewable generation, energy storage and large-load projects are stalled in grid connection queues worldwide. Annual grid investment, currently around $400 billion, needs to increase by roughly 50% by 2030 to meet forecast electricity demand.

There is an important caveat here: a connection queue is not the same thing as a pipeline of projects certain to be built. Many projects are speculative or will eventually be withdrawn.

But that does not make the bottleneck imaginary.

It tells us something else instead: electricity infrastructure is increasingly becoming a constraint on both sides of the market – on companies trying to generate clean electricity and on companies trying to consume much more of it.

Grid capacity is becoming economic capacity.

The timing mismatch is becoming structural

The problem starts with speed.

Renewable generation can now be deployed comparatively quickly. The same is true of many of the new technologies consuming electricity.

According to the IEA, solar and wind projects can typically take one to five years to develop. EV charging infrastructure can take one to two. Data centres may take one to three.

Major grid infrastructure can take five to fifteen years.

Think about that mismatch.

We can build the asset demanding the electricity faster than we can build the infrastructure supplying it.

We can also build the asset generating the electricity faster than we can build the infrastructure carrying it away.

The grid gets squeezed in the middle.

AI is making one part of that problem particularly visible. The IEA now expects global data-centre electricity consumption to rise from around 485 TWh in 2025 to 950 TWh by 2030, with AI-focused facilities growing considerably faster than the data-centre sector overall.

But AI is only one source of new demand.

Transport is electrifying. Heat is electrifying. Parts of heavy industry are electrifying. Battery manufacturing, semiconductor production and other advanced industrial processes can require significant and highly concentrated power loads.

Electricity is moving from being one input into the economy to becoming a much larger part of its energy foundation.

And infrastructure built for yesterday’s demand patterns is being asked to support tomorrow’s.

Generation without delivery creates stranded value

This was the idea at the centre of my recent Climate Confident conversation with Niklas Persson, CEO of Grid Integration at Hitachi Energy.

Niklas’s point was deceptively simple: building more renewable generation without expanding and modernising the grid eventually means some of that electricity cannot be used.

It gets curtailed.

Australia offers a useful illustration.

The Australian Energy Market Operator reported that grid-scale solar in the National Electricity Market experienced average network-driven curtailment of 4.5% in 2024, while several individual solar farms saw curtailment above 25%. The worst congestion was concentrated in areas where networks originally built mainly to supply local electricity demand were now being asked to export substantial volumes of renewable generation.

That is an engineering problem.

It is also an economics problem.

Imagine doubling production at a factory while leaving the road outside unchanged, then discovering that trucks cannot move the additional output to customers.

Nobody would call that a manufacturing success.

They would call it a logistics failure.

Electricity has its own equivalent of production, routing, storage, capacity constraints, congestion and last-mile delivery.

The grid is the logistics system of the electrified economy.

More generation matters enormously. But production is only valuable when it can reach demand.

Power availability is becoming a site-selection issue

This is where grid infrastructure lands squarely in the boardroom.

The European Commission says grid connection queues now exist in at least 16 EU countries, with around 120 GW of mature renewable projects, including 1.5 million household installations, at risk of failing to receive timely grid access by 2030.

Across the US, Berkeley Lab recorded 2,061 GW of generation and storage capacity seeking transmission connection at the end of 2025.

Again, that number needs context. Berkeley Lab found that historically only a minority of queued capacity ultimately becomes operational, while 75% of capacity entering queues between 2000 and 2020 had been withdrawn by the end of 2025.

Yet another number is arguably more telling: for projects that actually reached commercial operation in 2025, the median journey through the connection process in regions with available data was more than five years.

For executives making capital decisions, that matters.

Imagine two regions competing for the same advanced manufacturing plant. Labour costs are comparable. Logistics are good. Incentives are attractive.

One can guarantee the required clean electrical capacity in two years.

The other says six.

Suddenly the substation belongs in the site-selection model alongside the motorway, port and workforce.

The same calculation applies to fleet depots, logistics hubs, semiconductor plants, battery factories and data centres.

Power availability becomes time-to-market.

And time-to-market becomes competitive advantage.

Build ahead of demand, or arrive behind it

That creates a difficult question for governments and regulators.

How much electricity infrastructure should be built before demand is certain?

Transmission assets are expensive and long-lived. Consumers ultimately carry much of their cost. Building infrastructure that goes badly underused is hardly good policy.

But waiting until every megawatt of future demand is contractually certain creates a different risk: the infrastructure may arrive years after the economic opportunity.

Europe is already moving towards what regulators call anticipatory investment – planning some network development against credible future demand rather than merely responding to connection applications already in the queue.

The scale is substantial. The European Commission estimates that around €730 billion of distribution investment and €477 billion of transmission investment could be required by 2040.

That does not mean writing blank cheques for grid operators.

Quite the opposite.

It means becoming better at forecasting where electricity demand and generation are likely to emerge, sharing the utilisation risk sensibly, improving spatial planning, coordinating infrastructure programmes and taking decisions early enough for the physical system to arrive when the economy needs it.

Niklas captured the other side of that risk wonderfully in our lightning round.

I asked him for the most expensive grid mistake.

His answer was two words:

“Not building.”

That is deliberately provocative. Overbuilding has costs.

But so does being late.

The second is simply harder to see because it appears as the factory that chooses another country, the charging depot delayed, the renewable project curtailed, or the investment that never reaches a final board paper.

Energy sovereignty may require electrical interdependence

There is another strategic contradiction hiding here.

Europe understandably wants greater energy security after seeing the consequences of excessive dependence on imported Russian fossil fuels.

But a more secure electrified system does not necessarily mean every country attempting to become an electrical island.

Quite possibly the opposite.

Denmark is a useful example. When wind generation exceeds domestic demand, interconnection allows electricity to move into neighbouring systems. Nordic hydropower can conserve water while wind is plentiful elsewhere. When Danish wind production falls, electricity can move back in.

Interconnection allows different weather patterns, generating technologies, storage resources and consumption profiles to complement one another.

It means countries do not each have to build an electricity system capable of independently surviving every conceivable peak.

That is not dependence in the old fossil-energy sense of relying heavily on one pipeline and one supplier.

It is a network of options.

And options are valuable in resilient systems.

HVDC is changing electricity’s geography

High-voltage direct current transmission pushes that logic further.

The engineering is sophisticated; the strategic idea is straightforward.

HVDC makes it practical to move very large quantities of electricity efficiently across long distances and through subsea cables. As a result, renewable resources no longer need to sit conveniently beside the consumers using them.

That changes electricity’s economic geography.

ELMED is one example.

The 600 MW electricity interconnector being developed between Tunisia and Italy will run more than 200 kilometres between Tunisia’s Cap Bon region and Sicily, creating the first direct electricity connection between the two countries.

Projects like this could eventually allow North African renewable resources to play a larger role in European electricity supply.

But they also require care.

Europe should not recreate an extractive energy relationship in which generating regions provide resources while capturing little of the economic value. Domestic electricity requirements, local jobs, skills, revenues, industrial development and infrastructure ownership all matter.

Done well, however, interconnection creates value in both directions.

Tunisia gains infrastructure, investment and access to a larger electricity market. Europe gains another source of power and additional system flexibility. Both gain another pathway for moving electricity when system conditions change.

The strategic asset is not simply the cable.

It is the optionality the cable creates.

We don’t have to wait for every new transmission line

There is a danger, though, in reducing the grid challenge to “build more wires”.

We certainly need more infrastructure.

But we also need to extract more value from infrastructure already built.

This may be one of the most commercially important findings in the IEA’s latest analysis.

It estimates that a combination of more flexible connection agreements, dynamic line ratings, advanced power-flow controls, reconductoring, voltage upgrades and related reforms could release enough capacity to connect 1,200-1,600 GW of advanced-stage projects currently caught in queues worldwide.

That is a huge number.

It reframes the problem.

Some of the grid constraint is physical infrastructure.

Some is how we operate that infrastructure.

Some is regulation.

Some is queue management.

Some is the assumption that every user must receive unrestricted access at every moment rather than accepting flexible connections in return for faster access.

Batteries, demand response, local generation and smarter siting add further flexibility.

So the choice is not “build grids” versus “optimise grids”.

We need both.

Electricity strategy belongs earlier in capital strategy

For senior business leaders, the practical conclusion is fairly stark.

Electricity can no longer be an engineering workstream added late in a major capital project.

For energy-intensive investments, management teams should understand available grid capacity, realistic connection dates, reinforcement requirements, power quality, potential curtailment or flexibility conditions, and the cost of alternative locations before committing heavily to a site.

That also means asking different questions.

Could storage reduce peak connection requirements?

Could a flexible connection bring the project online earlier?

Would locating closer to generation or existing network capacity materially alter project economics?

Could on-site generation cover part of the load?

What happens to the business case if the grid connection slips by three years?

Those are no longer niche energy questions.

They are investment questions.

The renewable generation revolution is well advanced. Solar, wind and batteries continue scaling because their economics increasingly work.

Now the infrastructure linking supply and demand has to catch up.

Niklas gave perhaps the best summary near the end of our conversation. Clean electricity only counts, he said, “when it can be brought to the consumers.”

Generation is not delivery.

And in an economy becoming steadily more dependent on electrons, delivery infrastructure determines where growth can occur.

That is why the grid can no longer remain invisible to boards, policymakers or investors.

The next battle for industrial competitiveness may not be won by the country offering the largest subsidy.

It may be won by the one that can say, with confidence:

We have the power. And we can connect you on time.

I explored the implications in much greater depth with Niklas Persson in the latest episode of Climate Confident, including interconnectors, HVDC, permitting, resilience and the difficult question of who gets priority when grid capacity is scarce.

Listen to Generation Is Not Delivery: The Grid Problem Behind Decarbonisation

Shortly afterwards, they invited me to verify the channel, so there is now a small verification tick beside the name as well.

Both are gratifying. I’m not going to pretend otherwise. The plaque is sitting beside me as I write this, and I’m rather pleased with it.

But the more interesting part of reaching 100,000 subscribers is not the plaque, the tick, or even the number itself.

It is what accumulated to get there.

Much of what I publish on YouTube consists of relatively long conversations about climate, energy, sustainability, technology and supply chains. I interview executives, scientists, entrepreneurs and specialists about subjects that can be technical, complicated and occasionally rather niche.

This is not an argument that serious content somehow shouldn’t work on YouTube. Plenty does. Nor does reaching 100,000 subscribers prove that there is a vast untapped audience waiting for every specialist with a microphone.

What my experience does suggest is something more useful: specialist knowledge and substantial audience reach are less incompatible than we sometimes assume.

But only if we stop thinking about content one piece at a time.

Consistency is not enough

One of the most common pieces of advice given to anyone trying to build an audience is to be consistent.

Keep publishing. Show up every week. Don’t stop.

There is truth in that, but I think it misses something important.

Consistency, by itself, is overrated.

Publishing something mediocre every Tuesday for five years simply leaves you with a very large collection of mediocre Tuesdays.

What matters is whether consistency leaves behind something that compounds.

I publish two podcasts. Resilient Supply Chain goes out on Mondays and Climate Confident on Wednesdays. Over the years, that has produced hundreds of interviews covering supply chains, technology, climate science, clean energy, emissions reduction, resilience and the messy realities of changing large organisations.

Each interview has an immediate life. It gets published, promoted, watched or listened to, discussed for a while, and eventually disappears from the current feed.

But it doesn’t disappear.

That distinction turns out to matter enormously.

A useful conversation recorded three years ago can still be discovered today. A viewer might arrive through one guest, then watch another interview on the same subject, then another. Someone searching for a specific technology or business problem can encounter work that I had mentally filed away years earlier.

Over time, individual pieces stop behaving quite so individually.

They become a catalogue.

Distribution is part of the work

Of course, a catalogue nobody discovers is not much of an asset.

This is another thing I have become more conscious of over time. Creating something good is not the same as ensuring that anyone encounters it.

Titles matter. Thumbnails matter. Openings matter. Searchability matters. So does promoting an interview through LinkedIn, newsletters, shorter video clips, blog posts and other channels.

That can make some people uncomfortable because packaging sounds dangerously close to marketing.

But I don’t think making an idea more discoverable requires trivialising it.

If I have a thoughtful 45-minute conversation with someone who knows far more about long-duration energy storage, supply-chain risk or industrial decarbonisation than most of us ever will, there is no virtue in giving it a title nobody wants to click.

The substance and the packaging perform different jobs.

One earns attention after the click. The other earns the click.

Both are part of publishing.

And the effect compounds. Every new piece of useful content provides another possible entrance to everything else.

You don’t need every piece to go viral

There is another lesson in the 100,000 number that I find particularly interesting.

It is easy to look at successful channels from the outside and imagine a succession of breakout hits.

That has not been my experience.

Some videos perform extremely well. Others don’t. Some get significant promotion; many rely heavily on organic discovery. And a 100,000-subscriber channel most certainly does not mean that 100,000 people eagerly appear every time I upload something. New videos can still begin with only a few hundred organic views.

Subscriber numbers and active attention are not the same thing.

That makes the cumulative effect more interesting, not less.

Virality is one mechanism for building an audience. Compounding is another.

A viral piece concentrates attention into a short period. A body of useful work distributes discovery across months and years. Somebody who has never heard of me can find one interview today, another next month and perhaps subscribe after the third.

There may never be a dramatic spike associated with that person.

But the relationship has changed.

For specialist publishing, I suspect we spend too much time looking for the spike and not enough thinking about the slope.

At some point, content becomes an asset

The biggest change in my own thinking has come from looking at what those hundreds of interviews have become collectively.

For years, I primarily thought of them as podcast episodes.

I increasingly don’t.

They are also a substantial collection of first-hand conversations with people working on some of the most consequential technological, industrial and environmental changes of our time.

That has opened possibilities I did not fully anticipate when I recorded them.

A peer-reviewed academic paper has already drawn on 52 interviews from the Resilient Supply Chain podcast. More recently, I have been systematically analysing the transcripts from both podcasts to identify recurring patterns, disagreements, operating problems and lessons that are difficult to see when each conversation is considered in isolation.

That work is feeding new research, articles and executive briefings.

Nothing about the original interviews changed.

What changed was the unit of analysis.

One interview is content.

Hundreds of related interviews can become a body of knowledge.

That, to me, is far more interesting than getting better at filling a content calendar.

It also raises a different question for organisations investing heavily in thought leadership.

Instead of asking, “What should we publish this week?”, perhaps the more valuable question is:

“What body of knowledge do we want to have created three years from now?”

Those lead to very different publishing strategies.

Authority compounds too

There is another reinforcing effect.

A substantial body of work gives a new reader somewhere to go next. It gives a prospective podcast guest evidence that they are not being invited onto something that appeared last Thursday. It gives conference organisers, clients and partners a way to see not merely what I claim to know, but what I have been discussing, questioning and exploring over a long period.

That reduces dependence on any individual post performing brilliantly.

It also creates a feedback loop.

Strong conversations help attract an audience. An audience makes the platform more attractive to strong guests. Stronger guests improve the conversations. The catalogue becomes more useful, which creates more routes for discovery.

None of those effects is automatic, and I certainly cannot isolate their relative contribution to reaching 100,000 subscribers.

But they are another reason I think audience size is better understood as a lagging indicator than as the asset itself.

The number tells you something happened.

The interesting question is what you built while it was happening.

There is no 100,000-subscriber formula

There is an obvious danger in writing an article like this: survivor bias.

I reached 100,000 subscribers, therefore I can explain how to reach 100,000 subscribers.

No.

Plenty of people publish consistently and never build a substantial audience. Subject matter matters. Quality matters. Guests matter. Timing matters. Packaging matters. Platform distribution matters. Promotion matters. Luck probably matters more than those of us who experience success generally care to admit.

And my own channel is hardly a controlled experiment.

I have changed formats, improved production, experimented with titles and thumbnails, promoted some content and published across multiple platforms. YouTube itself changes constantly. I cannot tell you precisely what percentage of those 100,000 subscriptions came from any particular decision.

Nor does a subscriber automatically translate into attention, influence or commercial value.

There is a chain:

attention → trust → authority → opportunity → outcome

Each step still has to be earned.

So my conclusion is not that everybody with specialist expertise should start a YouTube channel and keep uploading until a silver plaque arrives.

It is that organisations and experts may be undervaluing what sustained publishing can create when they judge every piece primarily by its immediate performance.

A post with modest reach can still add something valuable to a coherent body of work. A podcast episode can remain discoverable years after its publication date. An interview can later become evidence for research nobody envisaged when it was recorded.

The useful life of good content does not have to end when the promotion does.

The number is not the destination

I’m delighted to have the Silver Creator Award.

And there is something genuinely satisfying about knowing that more than 100,000 people have at some point encountered my work and made the deliberate decision that they would like to see more of it.

But I don’t think the most interesting objective now is 200,000.

The better question is what I can do with what has already accumulated.

Can those hundreds of conversations produce better research? Can they reveal patterns that individual interviews cannot? Can they help senior leaders make better decisions? Can old conversations become newly useful when viewed alongside newer ones? Can the next hundred interviews strengthen everything that came before them?

That is where my attention increasingly lies.

Because after years of publishing, I’m coming to think the most valuable thing I built was not a YouTube channel with 100,000 subscribers.

It was a body of work large enough to keep creating value long after I pressed publish.

Research the guest. Record the conversation. Edit it. Publish it. Promote it. Then move on to the next one.

The interviews remained available, of course, but the unit of value was always the episode.

That changed when five researchers: Stefan Seuring, Jannik Neuberger, Sharfah Ahmad Qazi, Lara Schilling and Andrea S. Patrucco — analysed 52 episodes of my Sustainable Supply Chain podcast for a peer-reviewed paper, Bridging Digital and Sustainable Supply Chains: Insights From a Podcast Analysis.

Their work examined how technologies including artificial intelligence, IoT, blockchain and cloud services were connected with sustainable supply-chain practices and outcomes.

I have written separately about what they found.

What interested me just as much was what their research implied.

They had treated podcast interviews not as media artefacts, but as qualitative research material.

That raised a larger question.

If meaningful insights could emerge from analysing 52 interviews together, what might be visible across the hundreds of other conversations I had recorded, but impossible to see when each episode was considered in isolation?

That question has changed how I think about the entire podcast archive.

From individual interviews to accumulated evidence

An individual interview can be valuable.

A supply-chain executive may explain why an AI deployment struggled. A technology provider may describe how a particular problem was solved. An energy specialist may outline how regulation is affecting investment. A sustainability leader may describe the operational consequences of new reporting requirements.

But viewed independently, each conversation remains one perspective at one point in time.

The analytical possibilities change when hundreds of dated interviews can be examined together.

Patterns can be compared across industries.

Claims made by technology providers can be tested against the experiences of operators.

Recurring implementation problems can be traced across otherwise unrelated conversations.

Ideas that were discussed enthusiastically several years ago can be compared with what people say about them after real-world deployment.

Apparent consensus can be tested by deliberately looking for dissent.

And because every conversation took place at a particular moment, changes in language, priorities and confidence can be examined over time.

That is when an interview collection begins to become something more than a back catalogue.

It becomes a potential source of longitudinal qualitative evidence.

The challenge is not generating answers

The obvious way to analyse hundreds of transcripts today would be to give them to an AI system and ask for the major trends.

That can produce an impressive answer.

The difficulty is establishing whether the answer deserves to be believed.

A plausible synthesis can combine genuine patterns, isolated anecdotes and the model’s own assumptions into something that sounds much more certain than the underlying evidence warrants.

So the problem I have been trying to solve is not simply how to extract answers from the transcripts.

It is how to retain enough provenance that those answers can be challenged.

If an analysis suggests that poor data foundations repeatedly undermine supply-chain AI initiatives, I need to know where that conclusion came from.

Who described the problem?

What role were they in?

Which organisation were they speaking for?

When did the interview take place?

What was the surrounding context?

And where, exactly, in the conversation was the relevant passage?

Without that chain back to the original source, the system may be useful for brainstorming.

It is much less useful for research.

First, I needed the complete record

There was a practical problem before any of this analysis could work properly.

I did not have transcripts for every episode.

For more recent interviews, that was not an issue. As transcription technology improved, transcription became part of my normal editing workflow, so I already had text versions of a large proportion of the collection.

The earlier years were different. When many of those episodes were produced, accurate automated transcription was either unavailable to me, expensive, or not practical enough to make it part of the process.

That left a significant hole in the historical record.

The solution came, somewhat unexpectedly, from Buzzsprout, the podcast hosting platform I use.

On a recent episode of their Buzzcast podcast, the team mentioned that creators could download transcripts of their back catalogue. I contacted their support team because many of my older episodes predated the point at which I had started generating transcripts myself.

They were able to run a process across the back catalogue so that those earlier episodes were transcribed as well.

That mattered far more than simply making old episodes searchable.

It meant I could finally bring the full collection of more than 800 interviews across both of my podcasts, into the same research environment.

Without that completeness, longitudinal analysis would always carry an awkward qualification: entire periods of the podcasts would effectively be invisible to the system.

With the historical transcripts available, the archive could be treated as a much more coherent record.

Then the harder analytical work could begin.

Making hundreds of conversations searchable

The first practical step is to divide long transcripts into smaller, usable passages.

An hour-long interview is too broad a unit for precise analysis. At the same time, fragments that are too small lose the context needed to understand what the speaker meant.

The goal is therefore to create passages that are specific enough to retrieve but substantial enough to preserve the argument.

Each passage remains connected to its source information: the episode, guest, date, speaker and, where available, timestamp.

That source connection is crucial.

It means analysis can move from hundreds of interviews to a specific piece of evidence and then back to the original conversation.

The next challenge is language.

Experts often describe the same underlying problem in completely different ways.

One person may talk about fragmented master data. Another may describe systems that cannot communicate. A third may explain that planners spend hours reconciling spreadsheets before they trust the numbers.

A conventional keyword search may treat those as unrelated statements.

Conceptually, they may be describing closely connected problems.

To address that, each passage can be represented mathematically according to its meaning. These representations, known as embeddings, make it possible to search for conceptually similar passages even when the vocabulary differs.

That makes a much larger proportion of the interview collection discoverable.

But it introduces another risk.

Retrieval is not evidence

Finding relevant passages is not the same as establishing a conclusion.

This distinction is fundamental.

Suppose a search retrieves 20 passages supporting one explanation and three that contradict it.

It would be tempting to interpret that as strong evidence that the first view dominates.

It is not.

The number of passages returned can be affected by the wording of the search, the structure of the transcripts, the retrieval method, repetition within individual interviews and the composition of the guest base.

Search frequency is not prevalence.

Ranking is not representativeness.

Model confidence is not evidence strength.

The retrieval system therefore has a narrower role: it helps locate material that deserves examination.

The analytical work begins after that.

Trying to disprove the finding

One of the most important parts of the process is deliberately searching for evidence that challenges an emerging conclusion.

Suppose several interviews suggest that supply-chain AI initiatives struggle to progress beyond pilots because the underlying data environment is fragmented.

A weak research process would continue searching for more examples of poor data.

A stronger one asks what would make that interpretation wrong.

Are there successful deployments that operated despite weak data?

Do practitioners identify a different constraint?

Do technology providers and operators describe the same failure differently?

Does the pattern persist across industries?

Has the explanation changed over time?

Is the apparent finding being driven disproportionately by a small number of guests or episodes?

The purpose is not to manufacture balance where none exists.

It is to understand how much weight a conclusion can legitimately carry.

Some patterns may be strongly supported across unrelated conversations.

Others may be suggestive but incomplete.

Some may turn out to depend on a particular sector, technology or type of contributor.

And sometimes the appropriate conclusion is simply that the available interviews do not provide enough evidence.

That is preferable to producing a confident answer that the material cannot support.

The importance of returning to the source

AI is extremely useful in this process because it can help interrogate a collection far larger than any individual could reliably hold in memory.

But I do not want the AI system itself to become the authority.

The authority remains the underlying evidence.

If a finding matters, it should be possible to identify the people whose accounts support it, the interviews in which they made those observations, the circumstances they were discussing and the passages on which the interpretation rests.

Contradictory evidence should be equally visible.

That is why timestamps and source metadata matter.

The goal is not to arrive at statements such as “the AI found that…”

It is to be able to say:

This pattern appears across these interviews.

These contributors describe the mechanism in these terms.

These other interviews challenge or qualify it.

And these are the limits of what the evidence allows us to conclude.

That is the methodological principle behind the Executive Briefings I have begun publishing from the interview collection.

A research base that keeps changing

There is another characteristic that makes this collection unusual.

It never really closes.

I continue to publish new interviews every week across my Resilient Supply Chain and Climate Confident podcasts.

Each new conversation becomes another dated contribution to the evidence base.

That creates the possibility of revisiting earlier conclusions rather than treating them as permanent.

A technology discussed largely in terms of promise several years ago may later be discussed in terms of deployment problems, economics, organisational resistance or measurable operating performance.

Themes can emerge, strengthen, weaken or change character.

The evolution of the supply-chain podcast itself reflects some of those shifts. It began as Digital Supply Chain, became Sustainable Supply Chain as decarbonisation and responsible sourcing moved closer to the centre of business strategy, and later became Resilient Supply Chain as disruption, geopolitical risk and operational continuity became harder to treat as separate concerns.

Those changes reflect both shifts in the business environment and my own editorial focus, which is important context when interpreting trends across the archive.

Differences between rhetoric and experience can become visible.

Predictions can eventually be compared with outcomes.

Time becomes part of the analysis.

And the relationship can operate in both directions.

Research into earlier interviews can expose contradictions, gaps and unresolved questions. Those gaps can then influence what I ask future guests.

The interviews produce research.

The research identifies missing evidence.

That missing evidence produces better questions.

And the resulting interviews strengthen the research base.

That feedback loop may ultimately prove more important than the underlying technology.

What this research cannot tell us

There is an important boundary around all of this.

The podcast collection is not a representative sample of the supply-chain, climate or energy industries.

Guests are selected rather than randomly sampled.

Editorial priorities influence which topics receive attention.

Some sectors, technologies and types of organisation are better represented than others.

Technology vendors feature prominently.

Public interviews may also be more likely to contain successful implementations than failed ones.

The academic researchers who analysed the 52 Sustainable Supply Chain episodes encountered a similar limitation. They noted that the material contained primarily positive accounts of digital technology use, making it difficult to assess unintended consequences and what they described as the darker side of digitalisation.

Those limitations do not invalidate the material.

They simply define what can legitimately be inferred from it.

The collection should not be used to claim that “most companies” believe something simply because a theme appears frequently.

It is much better suited to identifying recurring mechanisms, disagreements, implementation experiences, emerging weak signals, changing narratives and questions that warrant further investigation.

It is qualitative evidence, not an industry survey.

That distinction matters.

From content archive to research infrastructure

This work is what led me to establish the Research section of this site and begin producing Executive Briefings based on structured analysis across multiple interviews.

The objective is not to summarise old podcast episodes more efficiently.

It is to ask questions that individual interviews cannot answer.

Why do apparently successful AI pilots fail to become durable operating capabilities?

Where do technology providers and practitioners describe the same problem differently?

Which operational obstacles recur despite successive generations of technology?

Which widely accepted assumptions have surprisingly little evidence behind them?

How do priorities change as technologies move from experimentation towards deployment?

And what do the interviews tell us now that they could not have told us when they were originally published?

There are now more than 800 episodes across the wider podcast collection, with new interviews being added every week.

For years, I thought I was building a back catalogue.

The academic paper forced me to recognise something else.

I had also been accumulating a dated record of how hundreds of senior business leaders working inside major technological, operational and environmental transitions understood those changes while they were happening.

Making that record usable required more than AI. It required complete source material, reliable provenance and the ability to return every important finding to the people and conversations behind it.

The opportunity now is to stop treating that record simply as published content.

And start asking what it can teach us.

It is seven o’clock on a cold, windless evening: demand is still high, solar output has disappeared, an interconnector is constrained and every available source of flexibility is suddenly valuable.

That hour may last four hours. It may last forty, or become a multi-day weather event.

We often discuss energy storage as though it were a single answer to all of these problems. It is not. A battery that smooths short fluctuations, a system that shifts solar electricity into the evening, and an asset designed to cover several windless days provide fundamentally different services.

Yet electricity markets, procurement programmes and investment models do not always recognise those differences. They can count megawatts while overlooking time.

That is the central finding of my new executive briefing, The missing market for time, based on a structured review of 50 interviews from my Climate Confident and associated podcast archives.

The conclusion is not that technology no longer matters. It plainly does. The more interesting conclusion is that long-duration energy storage is not waiting for one decisive invention. In many cases, it is waiting for power systems to define the service they need, value it properly, contract it over an investable period and create a credible route from procurement to operation.

Battery storage is booming. Long duration is not

Storage is already one of the fastest-moving parts of the energy economy. According to the International Energy Agency’s Global Energy Review 2026, 108 GW of battery storage was deployed worldwide in 2025, 40% more than in 2024. Installed capacity is now eleven times higher than in 2021.

That is amazing progress. It can also obscure the problem.

The IEA says most projects still cluster around two hours, although durations are beginning to lengthen. Those systems can shift solar output, provide balancing services and reduce peaks. But success in short-duration batteries does not mean the market for assets covering ten hours, several days or longer is advancing at the same rate.

The scale required across the broader storage category is substantial. The IEA estimates that global capacity must reach 1,500 GW by 2030 to support the tripling of renewable generation while maintaining electricity security. Batteries provide most of that growth in its scenario. This is not, however, a 1,500 GW target for long-duration storage.

Even the term “long-duration” is unsettled. A US National Renewable Energy Laboratory review found definitions ranging from more than two hours to seasonal storage, with ten hours a common threshold. Buyers can discuss Long Duration Energy Storage (LDES) without agreeing on the problem it must solve.

Start with the uncovered period. Model what remains after transmission, interconnection, demand response, renewable diversity, short-duration storage and firm generation have all been considered. Then specify the necessary duration, location, response speed, cycling pattern and reliability.

The correct unit of analysis is not “a storage project”. It is the system condition that must be covered.

Electricity markets are good at pricing energy. Time is harder

In my recent Climate Confident conversation with Julia Souder, CEO of the Long Duration Energy Storage Council, her central argument was that markets still struggle to price time.

Wholesale markets can put a visible price on electricity delivered now and reward short-term balancing or reserve. Fewer provide stable, duration-sensitive revenues for dependable availability across a prolonged period of system stress.

An LDES asset can potentially provide energy arbitrage, capacity, congestion relief and ancillary services. On a presentation slide, these can be assembled into an attractive “revenue stack”. In the real market, however, the asset may not have access to every service, may not receive payment from every beneficiary, or may lack sufficient certainty about future revenues to support financing.

Value exists. A bankable cash flow may not.

The US Department of Energy’s 2025 commercial-liftoff analysis reaches a similar conclusion. It considers capacity markets, long-term bilateral contracts, cap-and-floor mechanisms, targeted tenders and more transparent system modelling among the potential routes to scale.

The striking point is not that one design has won. It is that technology improvement and market design must proceed together.

Targets do not finance projects. Contracts do

Governments and utilities increasingly recognise the need for storage. Recognition is welcome, but a target is not a transaction.

A developer cannot finance a project with a press release announcing a future gigawatt goal. Investors need to understand construction risk, performance, counterparties and lifetime revenues.

Procurement design therefore matters enormously.

If a tender is written around a familiar short-duration product, longer-duration technologies may be excluded before their system value is compared. If selection is dominated by initial capital cost, an asset with a longer useful life, more cycles or lower replacement costs can appear uncompetitive even when it offers a better lifecycle result.

Procurement should be technology-neutral and performance-specific. Buyers should define the service, require evidence, allocate risk explicitly and allow qualified resources to compete.

Where merchant markets cannot yet support financing, competitive long-term contracts may be needed to bridge the gap. These should retain demanding availability and performance obligations. Public support should help discover prices, build operating evidence and reduce risk, not guarantee every project a return.

The goal is a repeatable asset class, not an endless parade of demonstrations.

A signed contract is not an operating asset

Grid connection queues are long. Permitting can take years. New designs may depend on immature supply chains. Utilities demand exacting reliability and service arrangements, often before manufacturers have accumulated enough production and operating evidence.

Technologies bring different constraints. Compressed-air storage has site requirements. Flow batteries trade energy density for long cycle life. Hydrogen incurs substantial losses when electricity is converted to hydrogen and back. Thermal storage can be compelling for industrial heat, but depends on the duty and integration pathway.

That variety is a strength. It also makes the idea of declaring “LDES is ready” or “LDES is not ready” largely meaningless.

Ready for what, where, for how long and against which alternatives?

The archive evidence supports readiness in the plural. Some technologies are commercially credible for defined applications now. Others still need cost reductions, operational proof or supply-chain development. None should be selected through a technology beauty contest detached from the system requirement.

Five decisions that would move the market

For senior energy leaders and policymakers, the practical agenda is clearer than the technology debate suggests.

First, model the residual flexibility gap. Identify the hours or days that remain difficult after other resources have been included. Do not begin with a preferred storage chemistry.

Second, specify the service in time. Duration, location, availability, recovery time and cycling requirements should appear in operational language that procurement teams and investors can use.

Third, create a route to investability. Targets must become competitive tenders and contracts with sufficient revenue certainty, measurable performance and a credible allocation of risk.

Fourth, make deliverability an investment gate. Grid access, planning permission, manufacturing capacity, testing, warranties, skilled labour and long-term service should be assessed before an award, not left as implementation details.

Fifth, govern a portfolio rather than a favourite technology. Storage must be compared with transmission, flexible demand, interconnection, firm clean generation and renewable overbuild. The objective is not to maximise storage. It is to deliver the required reliability at the lowest credible system cost.

The market is moving, but not evenly

The 108 GW of battery capacity added in 2025 shows that storage can scale rapidly once technologies, revenues and customer needs align. Around 80% of those additions were utility-scale, according to the IEA. Durations are gradually increasing too, with more projects reaching four hours or beyond.

But four hours is not forty. The commercial momentum behind lithium-ion systems should not be mistaken for proof that every longer-duration pathway has crossed the same threshold. Interviews in my archive point to larger projects and accumulating performance data, but much of that evidence comes from technology providers and has not been independently audited.

Industrial applications are broadening the opportunity too. My conversation with Rondo Energy CEO John O’Donnell explored electric thermal storage as a way to convert renewable electricity into dependable industrial heat. That is not interchangeable with every grid-storage application, but it illustrates a crucial point: storage becomes valuable when it is designed around a specific service and connected to a real operating need.

Moving energy through time, and through space

Return to that difficult hour after sunset.

Storage can move plentiful electricity from an earlier period into it. But storage cannot solve every constraint. Sometimes the required electricity exists at the same moment in another region. The obstacle is the network between them.

This is the complementary infrastructure challenge at the heart of the energy transition:

Storage moves energy through time. Grids move energy through space.

The IEA estimates that meeting the global renewable goals agreed at COP28 requires not only 1,500 GW of storage by 2030, but also 25 million kilometres of new or modernised electricity grids. If grids and storage lag behind generation, its modelling shows more curtailment, higher prices and almost 40% more coal generation than under full implementation.

I will return to that second half of the equation in another executive briefing in the coming weeks, examining what prevents electricity networks from expanding and adapting at the pace the transition requires.

For now, the storage lesson is direct. We do not merely need more capacity. We need power markets and institutions capable of identifying which hours matter, paying for dependable service and turning technical capability into assets that can be financed, connected and operated.

That is the focus of my new executive briefing, The missing market for time. Drawing on evidence from 50 interviews, it sets out a practical decision framework for leaders responsible for energy planning, procurement, investment and delivery.

Read The missing market for time.

The next phase of the energy transition will be defined not simply by how much clean electricity we produce, but by whether we can deliver it at the right place and the right time.

The model produces a better forecast. The dashboard looks polished. The project team demonstrates recommendations that planners could not have generated as quickly on their own.

Approval follows. So does talk of rolling it out across the network.

Six months later, the model is still running, but the planners are back in spreadsheets. Recommendations arrive too late to affect the decision. One site trusts the output; another ignores it. When the data deteriorate, nobody owns the repair. When a recommendation is wrong, nobody is certain who owns the consequence.

The technology worked.

The operating model did not.

That distinction is the subject of my new executive brief, Why supply-chain AI pilots fail to deliver lasting value.

The report draws on a structured analysis of 43 interviews from my Resilient Supply Chain podcast. Those conversations bring together people designing, selling, implementing and working with supply-chain technology. Rather than relying on one survey or a handful of headline case studies, I compared their accounts to identify recurring operating problems, successful deployments, contradictions and areas where the evidence remains weak.

The central conclusion is simple:

The unit of scale is not the model. It is the operating capability surrounding it.

The pilot passed the wrong test

A proof of concept can establish whether a model is capable of producing a useful forecast, recommendation, classification or alert.

That is valuable. It is also incomplete.

Pilots operate under unusually favourable conditions. The scope is controlled. Data are selected and prepared. A small team pays close attention. Exceptions can be handled manually. Project sponsors are available to settle disagreements. If an output looks wrong, somebody investigates.

Production removes that protective layer.

Data continue to arrive from ERP and planning systems, spreadsheets, equipment, suppliers and logistics partners. Users have competing priorities. Local conditions differ. Decisions are time-sensitive. Experienced people change roles. Integrations fail. Models encounter events poorly represented, or entirely absent, in their historical data.

At that point, model performance is only one part of the investment case.

The harder question is whether the organisation can turn the output into a better operational decision, consistently enough to justify the full cost of doing so.

That full cost includes much more than the model or software licence. It includes integration, sensing, data maintenance, governance, support, process redesign, training, exception handling and provider continuity.

A technically successful pilot can therefore conceal a commercially weak production proposition.

Value is lost at the hand-offs

Across the interviews, four mechanisms received strong support: production data, connection to execution, decision authority and work design.

They should not be treated as four independent items on an AI-readiness checklist. They form a chain.

Break any critical link and value can disappear.

From prepared data to production data

A pilot can be built around a curated dataset. Production depends on information that continues to arrive at the required quality, frequency and speed.

That changes the nature of the data problem.

It is no longer a one-off cleansing exercise. It is an operating responsibility with named owners, maintenance rules, latency requirements, external dependencies and recovery procedures.

In one interview, Andy Kohm of SCIP explains how contradictory and poorly maintained source data can cause AI to accelerate the wrong decisions.

The danger is not simply that the model stops working. It may continue to produce confident, plausible outputs while the quality of the underlying signal declines.

That is harder to detect, and potentially more expensive.

From insight to action

Even an accurate recommendation has no economic value until it changes a real decision.

This sounds obvious. In practice, it is where many analytical systems become detached from operations.

Spencer Penn of Lightsource.ai describes sourcing decisions that may be completed through email and spreadsheets before the ERP record is updated. Intelligence connected only to the formal system of record can arrive after the consequential choice has already been made.

The model has an insight.

The workflow has moved on.

Similar failures occur when an alert reaches somebody who cannot act, when a recommendation is not integrated with the system of execution or when a standard design collides with the realities of a particular warehouse, factory or transport operation.

The production test must therefore include the entire action path: where the output appears, who receives it, which system executes it, how quickly action must follow and how the result returns as feedback.

From recommendation to authority

The next hand-off is organisational.

What may the AI observe? What may it recommend? What may it execute? When must a person intervene? Who may override the system? Who owns the result?

These are operating-model decisions, not technical settings.

Simon Bezrukov of Bristlecone distinguishes between automating the administration around a decision and owning its consequences. A system might gather missing information, open a ticket, propose a revised plan or select from an approved set of responses. That does not mean it should receive unrestricted authority over every decision.

“Human in the loop” does not resolve the issue on its own.

A nominal approval step can add delay without adding judgement. Equally, removing people too quickly can leave nobody capable of recognising a change in conditions, challenging a plausible error or taking responsibility for an exception.

Human involvement should reflect consequence, reversibility, confidence and the maturity of the deployment.

Authority can expand as evidence accumulates.

From adoption to work design

When people reject a new system, the explanation is often “resistance to change”.

Sometimes it is.

But that phrase can also conceal a design failure.

The software buyer may not be the person expected to use the output. A standard process may ignore legitimate site differences. An alert may add work without removing an existing task. Automation may eliminate the routine activities through which less experienced employees acquire judgement.

Training cannot compensate for a poorly designed job.

A production deployment must be developed with the people who will use and supervise it. It must be tested under realistic pressure, not only in a demonstration environment. That means rehearsing exceptions, degraded data, integration failures, provider outages and the moments when a person must take control.

The objective is not to persuade users to accept the technology.

It is to design a better way of working.

The successful cases sharpen the lesson

A failure-only account would be just as misleading as uncritical enthusiasm.

The interviews also contain examples of AI and automation producing operational value. These cases challenge two common assumptions.

The first is that an organisation must perfect its entire data estate before it can begin. It does not.

Focused modelling, simulation and an 80/20 approach can support a bounded decision without cleansing every field in every system.

The second is that every AI-supported action requires permanent human approval. It does not.

Opening a service ticket is not the same as committing inventory. Retrieving missing information is not the same as selecting a supplier. Adjusting a low-risk parameter within an approved range is not the same as controlling a safety-critical process.

The right degree of autonomy depends on the decision.

Some contributors also report material operational outcomes. Penn, for example, describes an unnamed automotive customer whose platform-supported categories completed sourcing 25% faster and experienced 37% less post-award cost creep than categories managed outside the platform.

That is a provider-reported comparison, not a controlled or independently audited study. It should not be treated as a universal benchmark.

It does, however, illustrate the architecture of a scaling case: a defined decision, operational integration and measures tied to the outcome.

The counterexamples do not weaken the four findings. They make them more precise.

The lesson is not to wait for perfect conditions.

It is to choose a consequential but bounded decision, test the complete operating chain and expand only when the evidence supports expansion.

Stop approving isolated technology experiments

Before approving another AI pilot, leaders should be able to answer eight questions:

1. Which consequential operational decision or workflow are we trying to improve?
2. Who owns that decision and its outcome?
3. What is the current performance baseline?
4. Can the required data be maintained under production conditions?
5. How will the output reach the person or system capable of acting?
6. What may the AI recommend or execute?
7. What happens when the data, model, integration or provider fails?
8. Which thresholds will cause us to scale, revise or stop?

Together, those answers form a bounded production hypothesis.

That is a better basis for investment than a loosely defined technology experiment. It forces the team to test whether the organisation can operate the capability, not merely whether the model can generate an impressive output.

It also creates a legitimate way for a pilot to succeed without being scaled unchanged.

A well-designed experiment may show that the expected value is not there. It may reveal an integration cost that changes the economics. It may demonstrate that process redesign would solve more of the problem than machine learning.

That is useful knowledge.

A pilot fails when it neither improves a material outcome nor resolves an uncertainty worth funding.

How the research was developed

I created the report by returning to recent Resilient Supply Chain interviews and retesting an earlier set of conclusions rather than assuming they were correct.

The analysis examined 43 interviews and 154 relevant transcript sections published between 3 November 2025 and 31 August 2026. It deliberately searched for evidence that challenged the emerging argument: successful deployments, alternative explanations, contradictory experiences and differences between what technology providers promise and what organisations report in practice.

This matters because retrieval is not proof. Finding several people making similar claims does not establish how common a problem is across the industry. Nor does a persuasive customer story become an independent benchmark simply because it includes an impressive number.

The strength of the research lies elsewhere.

It brings together detailed accounts from people working across supply-chain technology and operations, compares the mechanisms they describe and tests whether the initial explanation survives contact with exceptions and disagreement.

Four connected mechanisms emerged with strong support: production data, connection to execution, designed authority and redesigned work.

A fifth finding, value discipline, is suggestive rather than equally established. Several contributors argue, persuasively, that AI initiatives should begin with a valuable business problem and measurable baseline. But the interviews do not contain enough retrospective evidence from cancelled pilots to establish weak ROI discipline as a primary cause as strongly as the other four.

The research is qualitative. The interview participants were not randomly selected, and technology providers and advisers are more heavily represented than operators. The findings cannot tell us what proportion of all supply-chain AI pilots fail for a particular reason.

They can do something more useful for an executive deciding whether to fund the next stage: identify recurring failure mechanisms, reveal important exceptions and sharpen the questions that should be answered before more capital and organisational attention are committed.

Ask the production question

Return to that successful pilot presentation.

The forecast may genuinely be better. The recommendation may be useful. The underlying technology may be excellent.

Lasting value will still depend on what happens next: maintaining the inputs, connecting the output to action, allocating authority, redesigning the work, managing exceptions and measuring whether performance improves after the project team steps away.

So the next time a pilot team presents an impressive result, do not ask only:

“Did the model work?”

Ask:

“Can the organisation operate it?”

That is the production question.

And it is the question that separates an interesting demonstration from a capability worth scaling.

Read the research and download the free executive brief.

No registration required.

An inbound shipment will miss its slot. The control tower identifies the delay, highlights the affected order and turns the screen red. By 9.15, the warning has reached procurement, logistics and the warehouse team.

Everyone can see the problem.

But who can change the dock schedule? Is there alternative stock? Can production be resequenced? Who is authorised to accept the additional transport cost? By the time those questions are answered, the useful response window has closed.

The technology worked. The operation still failed.

This is the visibility trap. Businesses have spent years treating end-to-end visibility as a route to resilience, often assuming that more data, more alerts and a more complete digital picture will produce a faster response. Yet information does not move a pallet, switch a supplier or protect a customer commitment. People and systems must convert it into a decision, and the physical operation must be able to carry that decision out.

That gap between seeing and responding is the subject of my new executive brief, Resilience Without Excess: Why Supply-Chain Visibility Still Fails to Create Resilience. It draws on a structured analysis of 43 interviews from the Resilient Supply Chain podcast archive, covering 78 relevant transcript sections and 160 validated episode-level claims.

The research points to a deceptively simple conclusion: resilience is not a dashboard feature. It is an operating capability.

Visibility has expanded. Resilience has not kept pace

The commercial urgency is difficult to miss. Business interruption, including supply-chain disruption, ranks third in the Allianz Risk Barometer 2026, after occupying either first or second place throughout the previous decade. Only 3% of respondents described their supply chains as “very resilient”.

Companies are not blind to the problem. They have invested in supplier mapping, planning platforms, control towers, sensors and predictive analytics. The view is improving, but unevenly. McKinsey’s 2025 survey of 100 global supply-chain leaders found that 95% had visibility into at least tier-one supplier risks. That figure fell to 42% for visibility extending to tier two or beyond.

Technology adoption reveals a similar divide between ambition and operational use. Three-quarters of respondents were planning, designing or piloting supply-chain AI applications; only 19% said they were deploying them at scale.

The problem, then, is not simply that organisations lack information. It is that the route from information to intervention is unreliable.

Across the podcast interviews, that route broke in five recurring places: the signal stopped at the screen; the digital record failed to reflect the physical operation; information stalled at a system or organisational boundary; governance left competing priorities unresolved; or the operation lacked the authority, resources and trust required to respond.

These are not five isolated technology defects. They are parts of one management problem.

A signal has value only while choices remain

Visibility is perishable.

A strategic network review may work with monthly data. A parcel jam, yard backlog or production constraint can deteriorate in minutes. In each case, information has value only if it arrives early enough for someone to choose a better outcome.

That makes “real time” a poor specification on its own. Second-by-second sensing is wasteful when the business cannot act until next week. A daily update is inadequate when the next hour determines whether a shipment makes its cutoff. Leaders need to define the decision window first, then determine what data speed and granularity it requires.

Accuracy matters just as much. A polished control tower cannot make contradictory part records, inconsistent supplier identifiers or stale inventory positions true. Nor can an advanced model reason about a pallet, product condition or machine event that the operation never captured.

The digital picture fails in two ways: the recorded data may be unreliable, or the necessary physical event may be absent. More sophisticated analytics can conceal the first problem and cannot repair the second.

This is why data work should not be dismissed as preliminary housekeeping. Establishing the source, ownership, quality, freshness and maintenance of a critical event is part of operational design. Without that discipline, visibility can become a more persuasive representation of conflicting information.

End to end is an agreement, not a screen

The phrase “end-to-end visibility” suggests a single vantage point over the entire supply chain. Real operations are less accommodating.

Orders, inventory, production, warehouse execution, yard movements, transport status, supplier capacity and risk data sit in different applications. They also sit in different companies, governed by different contracts, incentives and standards. One platform may describe its portion accurately while missing the dependency that determines the outcome.

The answer is not to pull every datum into one enormous repository. It is to identify the minimum reliable information that must cross each boundary for a critical decision to be made.

That requires explicit agreements. Who creates the event? Who assures it? How fresh must it be? Which organisation can use it? What happens when it is missing? Where does responsibility sit when a critical service has been outsourced to a provider whose own dependencies remain opaque?

This last question matters. A company can outsource software, logistics or data processing. It cannot outsource accountability for the disruption that follows when those services fail.

Visibility across organisational boundaries is therefore as much a commercial and governance challenge as a technical one. Integration architecture matters. So do data standards, partner obligations, portability, escalation rights and the willingness to engage suppliers rather than merely send them another questionnaire.

Better information does not settle a business trade-off

Even a shared, accurate view leaves a harder question: what should the organisation do?

Procurement may optimise purchase price. Finance protects cash and working capital. Operations prioritises throughput. Customer teams protect service. A visibility platform can expose the tension among those objectives, but it cannot decide how the enterprise should value them.

More information can even deepen fragmentation if each function uses it to optimise its own scorecard faster.

Resilience requires choices that may look inefficient through a narrow cost lens. Extra capacity, alternative suppliers or inventory flexibility carry a visible price; the disruption they prevent remains hypothetical until it occurs. Senior leadership must decide how cost, cash, service and risk will be balanced before a time-critical alert forces the debate.

That means every important visibility product needs more than a sponsor. It needs a decision owner, defined escalation thresholds, agreed trade-offs and a forum capable of resolving conflict. A metric without a required response is reporting. A warning without authority is noise.

The final mile is physical, and human

Suppose the data is reliable, the boundary has been crossed and the decision is clear. The response can still fail.

A yard may need another driver. A warehouse may lack space. Production may require a configuration change. A supplier switch may depend on contractual approval. A site may operate differently from the process designed at headquarters. The user receiving the recommendation may not trust it, understand it or possess the authority to act.

These are not downstream implementation details. They determine whether the investment can alter performance.

The interviews repeatedly returned to local variation, frontline behaviour and deployable capacity. Technology must fit the operation as it exists while helping that operation improve. That requires training, configuration, clear decision rights and safe fallback procedures. It also requires testing under real conditions, where labour is constrained, data can fail and disruption rarely follows the demonstration script.

“Change management” is too often treated as communication after the technical work is complete. In a resilient operating model, adoption is part of the product specification from the beginning.

Design backwards from the consequence

The usual technology discussion starts with capability: What can the platform display? Which systems can it connect? Where can AI generate a prediction or recommendation?

Leaders should reverse the sequence.

Start with the consequence. Which service, cost, safety or recovery outcome deteriorates because the business reacts too late?

Name the decision. What must be decided, by whom and within what window?

Design the action path. Which person, workflow, partner, contract, vehicle, labour pool or inventory position can change the outcome? If there is no credible intervention, more visibility will mostly document failure in greater detail.

Then specify the minimum data contract: required events, granularity, freshness, assurance, system of record, maintenance owner and acceptable failure rate. Align the trade-off among cost, cash, service and risk. Finally, test the complete loop in operational time.

The success measure is not the number of connected sources or deployed dashboards. It is whether decisions became earlier, interventions were completed and operational outcomes improved.

This approach also gives AI a more useful role. Models can detect patterns, rank options and remove administrative friction at remarkable speed. But they cannot compensate for a missing physical event, an unresolved commercial trade-off or an unavailable truck. If the operating loop is broken, AI may generate recommendations faster than the organisation can absorb them.

From visibility to resilience

The encouraging signal is that the supply-chain conversation is maturing. Leaders are asking less about whether they can see an exception and more about whether they can act before it becomes a consequence. Attention is moving towards decision latency, deeper-tier relationships, data ownership, human accountability and operational execution.

That does not reduce the importance of visibility. It completes its purpose.

Connect the signal to a decision. Connect the decision to an owner. Connect the owner to an available action. Then verify that the action changed the outcome.

That is resilience without excess: not endless data, inventory or technology, but enough dependable information, authority and capacity to respond while meaningful choices remain.

The red alert at 9.07 should be the beginning of the response, not the end of the achievement.

My new executive brief, Resilience Without Excess: Why Supply-Chain Visibility Still Fails to Create Resilience, is available to read and download free: read the full research brief.

Sources

• Tom Raftery, Resilience Without Excess: Why Supply-Chain Visibility Still Fails to Create Resilience (2026): https://tomraftery.com/research/resilience-without-excess/
• Allianz Commercial, Allianz Risk Barometer 2026: https://commercial.allianz.com/content/dam/onemarketing/commercial/commercial/reports/allianz-risk-barometer-2026.pdf
• McKinsey & Company, Supply chain risk pulse 2025: Tariffs reshuffle global trade priorities (2 December 2025): https://www.mckinsey.com/capabilities/operations/our-insights/supply-chain-risk-survey

When I started recording conversations with supply-chain executives, technologists and practitioners, the aim was straightforward: understand what people close to the work were actually seeing, building and learning.

One episode became ten. Ten became a hundred. Eventually there was an archive large enough to contain something I had never set out to create: a dataset.

Researchers Stefan Seuring, Jannik Neuberger, Sharfah Ahmad Qazi, Lara Schilling and Andrea S. Patrucco have now analysed 52 of those interviews for a peer-reviewed paper in the International Journal of Physical Distribution & Logistics Management. 

The interesting part is not that a podcast became research material.

It is what they found inside the conversations.

And the strongest finding is a useful corrective to much of the technology rhetoric currently washing through boardrooms: technology is not the transformation.

What the researchers actually analysed

At the cut-off point in June 2024, the podcast archive contained 404 episodes. The researchers used a purposeful selection process to identify 52 conversations dealing substantively with both digital technology and supply-chain sustainability. 

They examined four technology groups, artificial intelligence, Internet of Things, blockchain and cloud services, against established sustainable supply-chain management practices and economic, environmental and social outcomes.

This was not simply a word count. Keyword analysis surfaced patterns; researchers then manually coded the transcripts for context and meaning and used contingency analysis to examine which technologies, practices and outcomes tended to appear together. A small practitioner “resonance” survey was added afterwards as a supplementary plausibility check. 

Podcast interviews are a slightly odd research source. That is partly their value. They are comparatively unscripted and practitioners can wander into examples, barriers and explanations that a tightly structured survey might never invite.

But they are curated media. Guests know they are speaking publicly. Hosts select them. Success stories travel better than post-mortems.

That limitation becomes important later. 

Technology creates capability. Management converts it into outcomes.

This is the conceptual heart of the paper.

The researchers did not find a simple line running from “deploy technology” to “achieve sustainability”. Practitioner narratives rarely described digital tools as directly producing better environmental, economic or social performance.

Instead, technologies enabled management practices, particularly sustainable risk management, proactivity and collaboration, and those practices were associated with sustainability outcomes. 

Sustainable risk management accounted for 48% of the practice-related coded segments, appearing across 87% of the selected episodes. Proactivity represented another 26%, while collaboration accounted for 15%. Broader strategic ideas such as organisational orientation and continuity of supplier relationships appeared less often as explicit technology links; the researchers interpret them more as enabling context. 

In practitioner language:

Technology creates capability. Management converts capability into outcomes.

AI can detect a supplier anomaly. It cannot decide what risk tolerance your business should accept.

IoT can tell you a refrigerated shipment spent too long above specification. It cannot repair a procurement process that rewards the cheapest supplier regardless of spoilage, emissions or service failure.

A blockchain ledger can make provenance harder to dispute. It cannot make a weak due-diligence process strong.

The implication for executives is that buying technology is the visible part. Data architecture, governance, decision rights, supplier engagement and process redesign are where the actual transformation happens.

What the technologies are actually good at

The paper is useful because it does not throw AI, IoT, blockchain and cloud into one digital-transformation bucket.

AI: less automation, more anticipation

AI was the most frequently discussed technology by some distance: 178 coded segments, 52% of all digital-technology mentions, across 29 of the 52 episodes. 

The use cases include predictive analytics, supplier-risk assessment, digital twins, predictive maintenance, scenario analysis and automated risk detection. But the pattern behind them is more interesting.

AI shifts the supply chain from explaining yesterday to estimating tomorrow.

A predictive-maintenance model can flag degradation before a line stops. A supplier-risk engine can combine signals that a category manager could never process manually. A digital twin can test routing, production or capacity choices before the downside appears in the physical network.

The researchers associate AI most strongly with proactivity and sustainable risk management. My interpretation is that anticipation may become AI’s most valuable supply-chain capability: buying decision time.

And decision time is resilience.

IoT: giving the physical supply chain a nervous system

If AI is the prediction layer, IoT is the sensing layer.

The interviews linked IoT to real-time visibility, automated tracking and tracing, condition monitoring, lifecycle assessment and product-level carbon measurement. Sensors turn physical events, temperature, location, vibration, equipment use, into data that systems and people can act upon. 

Better sensing can reduce spoilage, disruption losses and waste while improving service and asset utilisation. More importantly, it closes part of the gap between what the digital supply chain says is happening and what is physically happening.

No data architecture can compensate for a blind physical network.

Blockchain: narrower, and therefore more credible

Blockchain gets a more restrained verdict. That is healthy.

The paper does not support the old claim that blockchain will remake supply chains wholesale. Its strongest role is narrower: traceability, certification, due diligence, proof of origin, compliance and verification. 

The practitioner resonance survey was notably more cautious about blockchain than IoT or cloud. That does not make blockchain a failure. Technologies often become useful after expectations collapse from “this changes everything” to “this solves a specific expensive problem”.

A verification technology does not need to transform the enterprise. It needs to make verification better.

Cloud: important because it disappears

Cloud services are the least glamorous part of the story and arguably one of the most important.

The interviews describe cloud as infrastructure for real-time data sharing, collaborative forecasting, scenario analysis, visibility, predictive maintenance and coordination across dispersed organisations. 

Cloud is rarely the sustainability intervention itself. It is the layer that lets other interventions operate across sites, partners and systems.

That is a recurring pattern in mature technology. The most consequential infrastructure eventually becomes boring.

The unexpected result: social sustainability moved to the centre

Environmental sustainability dominated the explicit discussion, accounting for 58% of sustainability-outcome coded segments. Economic outcomes represented 22%. Social outcomes represented 20%. 

If you stopped at frequency counts, environmental performance would look like the obvious centre of gravity.

The contingency analysis told a more interesting story.

Social outcomes sat at the centre of many relationships between digital technologies and sustainability. AI, IoT and cloud services showed comparatively strong positive associations with social outcomes in the coded material, appearing alongside working conditions, human rights, supplier compliance, safety and ethical sourcing. 

The researchers are careful here, and we should be too: co-occurrence is not causality. The study does not prove that installing IoT improves human rights.

What it does show is that practitioners repeatedly connect digital visibility with the ability to see social risk.

That matters. Supply-chain digitalisation is becoming part of the evidence infrastructure through which a company can establish whether it knows what is happening beyond Tier 1.

That regulatory pressure has not disappeared, despite considerable EU simplification. Under the amended Corporate Sustainability Due Diligence Directive, very large companies within scope must identify and address actual and potential adverse human-rights and environmental impacts in their operations, subsidiaries and chains of activities. Following the 2026 Omnibus amendments, member states are due to apply the revised rules from July 2029. European Commission

The management question is shifting from “Do we have a supplier code of conduct?” to:

Can we produce credible evidence that the standards are being followed?

The part of the research I found most uncomfortable

The paper also points a finger at the dataset itself.

Most conversations are positive. Many guests are technology providers, 64% of the organisations represented in the sample. Failures, internal resistance, unintended consequences and negative impacts are underrepresented. 

That matters most around AI.

The selected interviews discussed predictive power, optimisation, safety and efficiency far more often than bias, energy consumption, workforce displacement, excessive monitoring or bad organisational consequences. The researchers explicitly warn that the dataset cannot provide a balanced account of the “dark side” of digitalisation. 

That is not a reason to dismiss the findings.

It is a reason to improve the questions.

I need to start asking more questions like:

• Where did this fail?
• What did you underestimate?
• What appeared in the business case only after implementation started?
• Who carries the downside?
• What resistance emerged from employees or suppliers?
• Which promised benefit never materialised?
• What trade-off became visible only after deployment?
• What would the sceptic in the room say?

Better questions produce better practitioner intelligence. Success stories tell us what is possible. Failure stories tell us what can prevent success, and that is often more useful when trying to replicate it..

What supply-chain leaders should do differently

The paper’s practical lesson is almost unfashionably sensible: start with the management problem, not the technology.

“Where can we use AI?” is usually a poor opening question.

Ask instead: Which risks are invisible? Which decisions arrive too late? Where is supplier data unreliable? Where does latency create cost, emissions or vulnerability? Which compliance processes remain manual? Where would prediction materially change the decision?

Then choose the technology.

The research also points to prerequisites that receive far less conference-stage attention than generative AI: integrated ERP foundations, accessible data, coherent architectures, governance and people capable of interpreting what the systems produce. 

Implementation should be sequential.

Instrument a critical flow with IoT before instrumenting everything. Apply predictive maintenance where downtime matters most. Use supplier-risk analytics where visibility is weak and exposure high. Use blockchain where provenance, certification or multi-party verification is genuinely difficult.

Then tie each project to an outcome somebody can measure: fewer disruptions, lower waste, better working conditions, reduced emissions, faster due diligence, improved service, less unplanned downtime.

Digital transformation should accumulate evidence, not slogans.

A podcast archive is also a record of an industry thinking

This is where the research changes how I think about the archive.

I started these conversations to explore where supply-chain technology, sustainability and operations were going. I did not set out to build a longitudinal record of practitioner thinking.

But that is increasingly what the archive is.

It contains changing attitudes to resilience, AI, Scope 3, procurement, supplier risk, geopolitics, reshoring, traceability and digital transformation across years in which those subjects moved from specialist conversations into executive priorities.

That does not mean the podcast has been “academically validated”. It means researchers judged a defined slice of the archive rich enough to analyse as secondary qualitative evidence, with explicit caveats about selection, bias and interpretation. 

The more interesting possibility is that podcast archives can become records of how an industry explains change to itself while that change is still happening.

Conversations accumulate

A podcast episode feels temporary.

One guest. One conversation. One week in a publishing schedule.

Then the episodes accumulate, and patterns begin to appear. Some become visible only when somebody steps back, codes the language, compares the themes and asks a different question of the archive.

For me, the deepest lesson from this research is not about podcasts at all. It is about transformation.

Technology gives organisations new ways to see, predict, verify and coordinate. None of that guarantees a better supply chain.

The outcomes still depend on management: what gets measured, what gets acted on, whose interests count, which trade-offs are accepted, and whether leaders ask what went wrong as carefully as they ask what worked.

They begin with a well.

A field.

A failed crop.

A neighbour asking why strawberries, irrigation, tourism, politics, aquifers, and Europe’s most important wetlands now seem to be fighting over the same shrinking resource. Around Doñana, this is no abstraction. In 2021, the European Court of Justice ruled that Spain had breached EU law because excessive groundwater extraction was damaging the protected area, and WWF has estimated that around 1,000 illegal wells and 3,000 hectares of illegal farms have contributed to unsustainable water use.

That is what resource constraint looks like when it stops being theory.

It becomes local.

It becomes political.

It becomes permission.

And that is precisely where artificial intelligence is heading.

AI is still spoken about in strangely weightless language. Models. Agents. Workflows. Productivity. Acceleration. The vocabulary is digital, frictionless, almost airborne. Yet the reality is anything but. AI sits in data centres. Data centres sit on land. They draw electricity from grids, water from local systems, hardware from strained supply chains, and legitimacy from communities, customers, regulators, investors, and employees.

That was the most important thread in my Climate Confident conversation with Sophia Mendelsohn, who leads SAP’s Global Sustainability Platform. Her argument was clear: AI needs sustainability. A company cannot achieve its AI objectives – or the growth those objectives are meant to deliver – without the discipline, data, and stakeholder experience that sustainability leaders hold.

That is not a moral flourish.

It is a business argument.

Because in 2026, AI is no longer merely a technology decision. It is an infrastructure decision. A procurement decision. A governance decision. A social licence decision. And, increasingly, a board decision.

The data says AI is becoming an industrial load

The International Energy Agency’s 2025 Energy and AI analysis estimates that global data centre electricity consumption will more than double to around 945 TWh by 2030 – slightly more than Japan’s current total electricity use. The same analysis projects data centre electricity demand to grow by around 15% per year between 2024 and 2030, more than four times faster than electricity demand growth from all other sectors combined.

The 2026 update sharpened the picture. The IEA reported that global data centre electricity demand grew 17% in 2025, while AI-focused data centres grew even faster, with electricity use from that subset rising 50%.

This is not a rounding error inside an IT budget.

It is a new industrial load arriving at speed, often in places where grids are already being asked to electrify transport, heat, buildings, and industry. The bottlenecks are physical: transformers, substations, switchgear, circuit breakers, land, permitting, water, grid queues, and skilled labour. Reuters reported in July 2026 that surging AI data centre demand is worsening shortages of critical electrical equipment in the US, with some lead times reaching up to 160 weeks. Wood Mackenzie expects US data centre capacity to rise from roughly 24 GW today to 110 GW by 2030.

That is the point many boardrooms still miss.

AI does not scale because a strategy deck says it should. It scales where power, water, hardware, permits, public consent, and procurement discipline allow it to scale.

Water is the most visceral example. The Uptime Institute rightly cautions against lazy comparisons between data centres and cities. In its 2024 Cooling System survey, only 14% of respondents with water-cooled data centres used more than 16 million US gallons, or around 60,000 cubic metres, per year.

But water is not a global average.

Water is local.

A litre used in a wet, cool region is not the same as a litre drawn from a stressed aquifer in a Mediterranean heatwave. That distinction matters because data centre impacts are felt at the level of specific grids, specific watersheds, specific communities, and specific planning authorities.

The question, therefore, is not “Are data centres efficient on average?”

The sharper question is: whose constraint does this facility worsen?

The implications are bigger than carbon

The easy version of this conversation is to say AI may increase emissions. True. But too narrow.

The real issue is that AI can collide with energy security, customer commitments, water politics, community trust, employee acceptance, industrial competitiveness, and long-term affordability. That makes sustainability a strategic discipline, not a reporting function.

For CEOs, the risk is not simply that AI adds tonnes of carbon to a Scope 3 inventory. The risk is that poorly governed AI creates hidden exposure across the business.

An AI contract may look clean on paper while masking grid congestion, water scarcity, weak supplier disclosure, volatile power prices, hardware replacement cycles, and reputational risk. A workload may appear cheap because the vendor has absorbed some costs for now. A procurement decision may look efficient because the sustainability questions were never asked.

That is not innovation.

That is deferred due diligence.

Sophia’s point is powerful because it reverses the usual caricature. Sustainability teams are often portrayed as the people who slow things down. In the AI infrastructure buildout, they may be the people who prevent programmes from stalling later – under the weight of energy constraints, community opposition, regulatory scrutiny, or customer pushback.

AI scales only where it is permitted to scale.

That permission comes from three constituencies.

First, the communities asked to host the infrastructure.

Second, the customers whose own public commitments may be contradicted by the technology they buy.

Third, the employees asked to adopt AI systems in their daily work.

That is social licence to operate. Mining, energy, chemicals, agriculture, and manufacturing have had to earn it for decades. Technology is now learning the lesson at uncomfortable speed.

The C-suite question, then, is no longer “How fast can we deploy AI?”

It is: “Can we deploy AI in a way that strengthens the business rather than quietly loading it with new resource, reputational, and regulatory risk?”

The strategies start before the contract is signed

The first move is procedural, but vital: put sustainability into AI procurement before contracts harden.

Not after deployment.

Not when the annual report is being assembled.

Not when a journalist asks where the data centre gets its water.

Before.

Every significant AI contract should ask where workloads run, how electricity is sourced, what cooling systems are used, what water risks exist, what hardware lifecycle assumptions sit behind the service, how e-waste is managed, how emissions are calculated, and whether the provider can report at a level useful for Scope 3 accounting.

The direction of travel in Europe is already clear. The EU Energy Efficiency Directive introduced monitoring and reporting obligations for data centres, and the European Commission’s database collects data relevant to energy performance and water footprint for facilities with significant energy consumption.

Second, boards need to broaden the AI ROI model.

Most AI business cases still over-index on speed, labour productivity, customer response times, automation, and margin improvement. Those are legitimate metrics. But they are incomplete.

A serious AI business case should also include energy exposure, water exposure, regulatory risk, vendor concentration, Scope 3 implications, grid delay risk, community acceptance, and operational resilience.

A narrow ROI model can make a fragile decision look precise.

Third, leaders should rank AI workloads by value density.

Not every prompt deserves expensive compute. Not every internal experiment deserves scale. Not every “AI-powered” feature creates strategic value. Some will be transformative. Some will be useful. Some will be computational theatre dressed up as digital leadership.

Token discipline is becoming financial discipline.

It is also becoming climate discipline.

Fourth, sustainability teams should use AI to change their own operating model. This is where the opportunity becomes genuinely interesting.

For years, Scope 3 work has depended on supplier questionnaires, partial disclosures, inconsistent spreadsheets, and the corporate equivalent of politely asking into the void. AI can shift that posture. Companies can model baselines, estimate product carbon footprints, flag anomalies, pressure-test assumptions, and then ask suppliers to confirm, correct, or improve the data.

That changes the power dynamic.

Sustainability moves from passive data collection to decision intelligence.

Finally, AI infrastructure should be designed around clean power, flexibility, and storage. The good news is that the clean technology stack is improving quickly. IRENA’s 2026 24/7 Renewables report estimates that four-hour utility-scale battery costs fell to around $140/kWh in 2025, close to 95% below 2010 levels.

That matters because AI loads do not have to be dumb loads. With better software, storage, grid signals, and contractual design, some workloads can become more flexible. Some can shift. Some can pair with clean power. Some can support grid stability rather than merely compete for capacity.

That is the strategic prize.

Not AI versus sustainability.

AI with sustainability built into the operating system of the business.

The signal of change is already visible

The transition is not hypothetical.

According to Ember’s Global Electricity Review 2026, solar met 75% of global electricity demand growth in 2025, while renewables supplied 33.8% of global power generation. Solar generation rose by a record 636 TWh in a single year, equivalent to roughly twice the UK’s annual electricity demand.

Clean power is scaling.

Storage is scaling.

Electrification is scaling.

And AI is now arriving as a major new claimant on that same electricity system.

This creates a fork in the road. Poorly governed, AI demand could extend grid bottlenecks, raise costs, intensify local water conflicts, and slow decarbonisation. Governed well, it could accelerate investment in renewables, storage, grid flexibility, efficiency, and better digital energy management.

That is why the sustainability function belongs in the AI conversation now.

Not as a compliance afterthought.

As a strategic partner.

The CEO should want sustainability in the room because sustainability leaders understand resource constraints. The CFO should want them there because hidden infrastructure risk becomes financial risk. The CIO should want them there because systems without trust fail adoption. The procurement leader should want them there because the leverage is greatest before the contract is signed. And the board should want them there because AI is becoming too material to treat as a siloed technology programme.

The companies that get this right will not simply have cleaner AI.

They will have more durable AI.

More investable AI.

More credible AI.

More useful AI.

And that brings us back to water.

A dry aquifer does not care about anyone’s innovation narrative. Neither does an overloaded grid. Neither does a community facing heat, rising bills, water stress, and another large infrastructure project at the edge of town.

Physical systems do not negotiate with PowerPoint.

AI’s promise is real. In medicine, science, engineering, climate modelling, materials discovery, logistics, and energy optimisation, it can be extraordinary. But the promise will only turn into durable business value if leaders treat AI as physical infrastructure, governed inside the constraints of power, water, land, materials, climate, and trust.

That is why my Climate Confident conversation with Sophia Mendelsohn matters. It is not another generic discussion about AI and sustainability. It is a board-level warning and a business opportunity.

The warning is this: treat AI as weightless software, and you will miss the risks until they become expensive.

The opportunity is better: bring sustainability into AI strategy now, and resource discipline becomes competitive advantage.

The next phase of AI will not be won by the organisations that deploy the most tools.

It will be won by the organisations that deploy the right AI, in the right places, for the right reasons, with the right infrastructure beneath it.

Listen to the full Climate Confident episode with Sophia Mendelsohn for the deeper conversation.

Photo credit dklaughman on Flickr

The myth says the work is the interview.

It is not.

The interview is the raw material. The real work is the system around it: how guests are selected, prepared, recorded, edited, packaged, hosted, published, and improved over time without depending on luck, adrenaline, or Sunday evening panic.

I have learned this over hundreds of episodes.

My Resilient Supply Chain podcast now has more than 510 episodes behind it. Climate Confident has passed 280. Together, they have generated more than 550,000 audio downloads and more than 600,000 YouTube views. Consistency at that scale is not an accident. It is a production system.

This is my current workflow. Current is doing quite a lot of work in that sentence.

It has changed repeatedly. It will change again. Podcasting is shifting from audio-first to multi-format media, and Apple’s move to support a stronger native video podcast experience inside Apple Podcasts is one more signal that production workflows cannot stand still. [1]

I think of the workflow as a media supply chain.

There are inputs, quality gates, dependencies, production steps, packaging, hosting, publishing, feedback loops, and a constant search for the next bottleneck.

For me, consistency is the anchor. Resilient Supply Chain goes live every Monday at 7am in Spain. Climate Confident goes live every Wednesday at 7am. The time itself is not sacred. The rhythm is. Listeners know when to expect a new episode. Guests and partners understand the cadence. I know what has to happen before each episode ships.

That predictability matters because podcasting has expanded well beyond audio. Edison Research reported in 2025 that 73% of Americans aged 12+ had consumed a podcast in audio or video format, 51% had watched one, and YouTube was the service used most often by U.S. weekly podcast listeners. [2]

A modern podcast is audio, video, search asset, credibility engine, and library of reusable thinking. Treat the recording as the finished product and most value quietly leaks away.

The workflow starts before anyone presses record.

I use online scheduling tools such as Cal.com and Calendly for intro and vetting calls. The call asks three questions: can the guest explain their ideas clearly, does the topic serve the audience, and is there enough substance?

If the fit is good, we schedule the recording. I ask the guest to send a few bullet points on topics they would be happy to discuss. These are not a script. Scripted interviews tend to sound like two corporate statements trapped in a lift. The bullets are a safety net if the conversation drifts.

Once the recording is booked, I add the recording link, platform details, guest notes, context, and audio/video guidance to the calendar invite. I ask guests to use a proper microphone if possible, avoid noisy rooms, wear headphones, and sit somewhere with decent light.

Some follow the advice. Some do not. This is why the workflow needs repair mechanisms.

My own hardware is intentionally stable: MacBook Pro, RodeCaster Duo interface, Shure SM7B microphone on a Rode PSA1 boom arm, Elgato Facecam, two Elgato Key Light Air lights, and Ecamm Live as my virtual camera.

This setup is about reliability, clarity, and repeatability. Gear should reduce friction. If it creates fresh anxiety before every recording, it is a tax on attention.

The setup has evolved too. I used to use my iPhone as a webcam through Apple’s Continuity Camera feature. It worked well. I now use an Elgato Facecam because it gives me more predictability and control.

That is the broader pattern.

I used Trint for transcription. Then Otter. Now transcription is handled inside Descript, alongside recording and editing. I used Audacity, then GarageBand, then Hindenburg for editing. Now I use Descript. For now.

That qualifier matters. This is not a love letter to a tool stack. Tools are useful until a better process, integration, or outcome appears. The loyalty is to the work, not the logo.

On recording day, I always have a warm-up conversation before the formal interview begins. Guests may be senior executives, founders, scientists, engineers, analysts, or policy specialists. They know their subject. That does not mean they are relaxed on camera. A few minutes of normal conversation changes the energy and lets me hear how they naturally speak before the recording starts.

I record using Descript Rooms. It captures local audio and video and keeps the session inside the same environment I use for editing. Descript says Rooms supports local recording with high-quality audio and video, including separate tracks, which gives more control later in production. [3]

After recording, the episode moves into editing in Descript.

This is one of the largest workflow changes I have made. Text-based editing changes the feel of audio and video production. Descript lets you edit media by editing the transcript, with the underlying audio or video changing as the text changes. [4]

That is powerful. It is not magic. And Descript is updated frequently. This is a double-edged sword. New features are added which can be good, but also existing functionality can move, behaviours can change, and things can occasionally break. Its AI functionality is useful too, but potentially eye-wateringly expensive if used carelessly as I found out to my cost when I allowed my older son to use it recently for a college project . AI is a tool. A blank cheque to a cloud model is not a production strategy.

My edit includes several stages. I correct the transcription, and remove filler words where appropriate. Then I create clips. I identify a cold opening, usually a short moment that captures the stakes of the conversation. I record the intro. I apply automatic multicam. I add the bumper and outro. Then I export subtitles, transcription, and the video.

From there, I prepare episode artwork in Canva, both for the podcast feed and for YouTube. Packaging matters. A strong conversation can still underperform if the title, artwork, or thumbnail fails to make the value clear.

Each episode then enters what I think of as the metadata factory: title, show notes, hashtags, guest profile, chapter markers, subtitles, transcript, episode number, and supporting details. Chapter markers are especially useful for long-form interviews. Podcast hosting company Buzzsprout describes them as labelled sections that help listeners move through an episode, preview what is ahead, or skip parts. [5]

Before publication, I email the guest to confirm the publication date and request a profile photo and bio if needed. This is ordinary admin, but ordinary admin is where many workflows become clogged. A missing bio, a late headshot, or a last-minute correction can slow everything else.

Once the video is exported, I run it through Auphonic to clean it up.

Auphonic has become hugely important because remote guest audio is inherently variable. I can control my microphone, room, and levels. I cannot control the guest’s laptop fan, kitchen acoustics, internet connection, keyboard noise, or the dog who has apparently chosen that exact moment to protest late capitalism.

Auphonic helps smooth some of that variation. Its published feature set includes noise and reverb reduction, intelligent levelling, filtering, AutoEQ, loudness processing, video support, metadata, and chapters. [6] For an interview-led podcast, that kind of post-production support is not cosmetic. It protects listenability.

After Auphonic finishes processing, I download the video and move into hosting and publishing.

Buzzsprout is the host for both of my podcasts. I have used other hosts in the past, including Libsyn and Podbean, and both have their place. Buzzsprout has become my preferred home because it is clean, practical, reliable, and backed by excellent support.

Support is easy to undervalue until something breaks.

When you publish every week, support is part of the resilience architecture. If an upload fails, a feed behaves strangely, or a metadata issue appears just before publication, you do not want to shout into a ticketing void. You want people who understand podcasting and respond with clarity.

Buzzsprout’s own customer-support podcast, Happy to Help, is a useful signal here. It is hosted by Priscilla Brooke, Buzzsprout’s Head of Podcaster Success, and focuses on making customer support better. [7] A company willing to talk publicly about support usually understands that support is part of the product.

For the podcast feed, I add the title, show notes, episode art, episode number, creator details, hashtags, processed video file, guest page, chapters, and custom transcription to Buzzsprout. Buzzsprout’s support for submitting video podcasts to Apple Podcasts also matters now that video is becoming a more central part of podcast distribution. [8]

Then I prepare YouTube separately: title, description, hashtags, thumbnail text, subtitles, playlist selection, and end screen. YouTube has its own discovery logic and viewer behaviour. A title that works for a podcast app may be too flat for YouTube. A podcast artwork tile may be too quiet as a video thumbnail.

Once the episode is packaged for both the podcast feed and YouTube, I schedule it for publication. In practice, I typically finish the edit the day before go-live, then schedule the episode to publish at the usual 7am slot.

Then, at 7am, it goes live. The waiting multitudes presumably abandon breakfast, sprint to their podcast apps, and download immediately. Or, in the slightly less theatrical version, the episode enters the world at the expected time, on the expected day, in the expected places.

That matters. Consistency builds trust quietly. No fireworks. No drama. Just showing up when people expect you to show up.

After that, the promotion process begins.

This post, though, is deliberately focused on production and publication. I am not covering promotion, monetisation, sponsorship, paid amplification, newsletter strategy, social distribution, or the broader business model here. Those are important, but they deserve separate posts.

For now, the lesson from hundreds of episodes is this: creativity scales only when operations carry it.

The warm-up conversation helps the guest. The bullet points protect the flow. The calendar invite reduces friction. The hardware stabilises quality. Descript consolidates recording, transcription, and editing. Auphonic reduces audio risk. Canva supports packaging. Buzzsprout hosts and supports the shows. YouTube requires its own publishing logic. Chapters, subtitles, transcripts, titles, artwork, descriptions, tags, and end screens turn one conversation into a structured media asset.

None of this is heroic. That is the point.

Heroics are a poor operating model. If a publishing system depends on repeated last-minute rescue missions, it is not a system. It is theatre with a deadline.

Podcasting is an operating system for trust.

And like any operating system, it needs maintenance, updates, guardrails, responsive partners, and a willingness to keep improving.

The surprise, for me, is that podcasting has taught me as much about supply chains as supply chains have taught me about podcasting. Flow matters. Bottlenecks matter. Quality control matters. Consistency matters. Support matters. Resilience is built before the disruption arrives.

This workflow will change again. It should. If you have suggestions, better tools, sharper practices, or lessons from your own production process, I would love to hear them. Continuous improvement is not a slogan. It is how the shows keep getting better.

Citations

[1] Apple announced a new HLS-enabled video podcast experience for Apple Podcasts in February 2026, including switching between watching and listening, horizontal display, and offline video downloads. 
[2] Edison Research’s Infinite Dial 2025 reported that 70% of Americans aged 12+ had listened to a podcast, 51% had watched one, 73% had consumed a podcast in audio or video format, and YouTube was the service used most often by U.S. weekly podcast listeners. 
[3] Descript describes Rooms as supporting local recording with high-quality audio and video, including separate tracks and up to 4K capture. 
[4] Descript describes its product as letting users record, transcribe, edit, and publish in one tool, with audio and video editing built around text. 
[5] Buzzsprout describes chapter markers as labelled sections that help listeners navigate, preview, or skip through podcast episodes. 
[6] Auphonic lists noise and reverb reduction, intelligent levelling, filtering, AutoEQ, loudness processing, video support, metadata, and chapters among its features. 
[7] Happy to Help is Buzzsprout’s customer-support podcast, hosted by Priscilla Brooke, Buzzsprout’s Head of Podcaster Success. 
[8] Buzzsprout provides guidance for submitting video podcasts to Apple Podcasts, and coverage of its launch notes that its video plans include transcripts. 

A weak ESG report.
A few awkward investor questions.
Maybe a raised eyebrow from procurement.
A consultant hired. A dashboard built. A target announced. Everyone went home feeling faintly virtuous.

That world is ending.

Increasingly, poor climate data will not just make a company look bad. It may make products harder to sell, harder to insure, harder to finance, harder to import, and harder to defend in front of regulators, customers, and investors.

This was the line that stayed with me from my recent Climate Confident conversation with Stephen Jamieson, Chief Marketing Officer for SAP Sustainability. Stephen made a deceptively simple point: if companies cannot meet emerging carbon data and product compliance conditions, they may simply not be able to put certain products on the market. He was talking about digital product passports, product carbon footprints, CBAM, metals, fashion, and the rising granularity of carbon data needed just to do business.

That is the shift.

Carbon data is moving from reporting obligation to commercial operating condition.

And for senior leaders, that changes everything.

The data says…

Let’s start with the policy stack, because this is not about one regulation, one disclosure framework, or one enthusiastic sustainability team trying to get procurement to return its calls.

The EU Carbon Border Adjustment Mechanism, or CBAM, entered its definitive regime on 1 January 2026. EU importers of covered goods now need to declare embedded emissions and surrender the corresponding number of CBAM certificates each year, with authorised declarant status becoming central to compliance.

Translation: carbon is becoming part of the cost of crossing a border.

Then there is the Ecodesign for Sustainable Products Regulation, or ESPR. The European Commission’s 2025-2030 working plan prioritises steel and aluminium, textiles, furniture, tyres, mattresses, and energy-related products for future ecodesign and energy labelling measures. Digital Product Passports sit inside this agenda, turning product sustainability from a nice brochure into structured, exchangeable product data. The QR code is not the strategy. The data behind it is.

Batteries show where this is heading. The EU Batteries Regulation takes a full life-cycle approach, covering sourcing, manufacturing, use, and recycling. From 2025, it gradually introduces carbon footprint declarations, performance classes, and maximum carbon footprint limits for electric vehicle batteries, light means of transport batteries, and rechargeable industrial batteries. It also provides for QR-code access to a digital battery passport with detailed product information.

Packaging is joining the queue too. The Packaging and Packaging Waste Regulation entered into force in February 2025 and will generally apply from August 2026. It covers all packaging and packaging waste, regardless of material or origin, and aims to make all packaging on the EU market recyclable in an economically viable way by 2030.

And this is broader than carbon. The EU Deforestation Regulation applies from 30 December 2026 for large and medium operators, requiring covered products to be deforestation-free. The EU Forced Labour Regulation will ban products made with forced labour from being sold in the EU market from 14 December 2027, applying to imports, EU-made products, and exports.

Different rules. Same direction.

Prove it.

At the same time, finance is tightening the lens. IFRS S2, effective for annual reporting periods beginning on or after 1 January 2024, requires companies to disclose climate-related risks and opportunities that could affect cash flows, access to finance, or cost of capital over the short, medium, or long term.

This is where the neat old boundary between “sustainability” and “business” collapses. Bad climate data is no longer just a disclosure gap. It is a pricing gap. A risk gap. A market-access gap.

The same pressure is visible in the voluntary and industry-led space. WBCSD’s Partnership for Carbon Transparency, or PACT, says supply chains cannot decarbonise without accurate and comparable product-level data, and its methodology is built around supplier-specific primary data for product carbon footprints.

Averages got us started.

They will not get us through this next phase.

The implications…

The implications are enormous, and not in the abstract “ESG is important” way that makes executives reach for their phones.

This is about product viability.

A company may still be able to manufacture a product. It may still have demand. It may still have a strong brand, loyal customers, and a procurement team that thinks it has things under control. But if the company cannot provide credible data on embedded carbon, recycled content, deforestation risk, packaging compliance, forced labour exposure, or product-level sustainability attributes, its commercial freedom narrows.

That is a different conversation.

It means sustainability data becomes part of revenue protection.

Procurement will need to know it.
Finance will need to trust it.
Operations will need to act on it.
Product teams will need to design with it.
AI systems will need to optimise around it.

This last point is crucial.

Stephen made a point in the episode that should make every AI-happy boardroom pause. Businesses are very good at measuring cost, revenue, utilisation, and finance. So AI will optimise for those high-quality inputs. If carbon, water, recycled content, resilience, and wider sustainability factors are not in the same decision environment, they risk becoming secondary, or worse, optimised against.

That is the danger.

AI will not wake up one morning with a conscience. It will optimise for what the system tells it to value.

So if sustainability data lives in a side platform, updated annually, disconnected from ERP, procurement, product lifecycle management, finance, and supply chain planning, then we should not be surprised when automated decisions accelerate the wrong outcomes.

Efficient nonsense is still nonsense.

There is an affordability angle here too. Clean technology is already pulling capital at scale. According to the IEA’s World Energy Investment 2025 report, global energy investment was set to reach USD 3.3 trillion in 2025, with around USD 2.2 trillion going to renewables, nuclear, grids, storage, low-emissions fuels, efficiency, and electrification, twice the USD 1.1 trillion going to oil, gas, and coal.

Capital is moving.

But capital follows confidence. And confidence follows data.

If a company can prove that its product has lower embedded emissions, lower exposure to carbon-intensive inputs, stronger traceability, and better compliance readiness, that is no longer just a sustainability claim. It is commercial differentiation.

The strategies…

So what should leaders do?

First, stop treating climate data as an annual reporting project. That model is too slow, too blunt, and too disconnected from the decisions that matter.

Start with a product exposure map. Which products rely on CBAM-covered materials? Which products fall into ESPR priority sectors? Which use packaging exposed to PPWR? Which contain batteries? Which touch commodities covered by deforestation rules? Which depend on suppliers in high-risk labour regions?

This is not box-ticking. It is revenue risk analysis.

Second, build what I would call a climate-data bill of materials. For priority products, companies need to know the materials, suppliers, production sites, energy inputs, packaging, recycled content, logistics, emissions factors, primary data availability, and evidence gaps.

Boring? Maybe.

Vital? Absolutely.

Third, put procurement in the middle of this. Supplier climate data cannot remain a favour requested politely by sustainability teams once a year. It needs to be built into supplier onboarding, contracts, scorecards, renewal criteria, and preferred supplier decisions.

Fourth, connect the systems.

This is where many organisations will struggle. Product carbon data needs to move across ERP, procurement, finance, product lifecycle management, compliance, supplier platforms, and digital product passport infrastructure. If the data cannot move, it cannot matter.

Fifth, be careful with AI. AI agents can help scale product carbon footprinting, compliance checks, emissions factor mapping, and documentation. Stephen described how AI capabilities could move companies from manually managing handfuls of emissions factors to scaling across tens of thousands of products. That is powerful.

But humans remain accountable. Stephen was clear on this too: AI agents can assist, but they cannot be accountable for decisions.

The strategy is not “let AI solve sustainability.”

The strategy is: give AI decision-grade sustainability data, strong governance, clear constraints, and human accountability.

The signal of change…

The transition is already happening.

CBAM is live. Product passports are being built. Battery rules are moving from aspiration to evidence. Packaging is becoming a compliance system. Deforestation-free proof is becoming a market requirement. Forced labour rules will put human rights evidence into product eligibility.

This is not a far-off future. It is a staggered rollout of commercial reality.

The signal of change is also visible in how business leaders now talk about climate. Stephen noted that climate has moved from a niche sustainability-team topic, through an era of ambition, into a phase focused on the economics of carbon, risk exposure, and investor relevance.

That tracks with what I am seeing too.

The serious conversations are no longer about whether sustainability matters. They are about how to make it operational. How to measure it properly. How to use it in procurement. How to price risk. How to make product-level carbon data trustworthy enough for investors, customers, regulators, and AI systems.

This is the real story.

Sustainability is leaving the PDF.

It is entering the transaction.

And that is good news, even if it feels uncomfortable. Because once climate data becomes part of how products are designed, sourced, financed, priced, and sold, it stops being a communications exercise and starts becoming a lever for emissions reduction.

Back to where we started.

A few years ago, bad carbon data meant a weak report.

Increasingly, bad carbon data may mean a weaker product, a weaker supplier position, a weaker investment case, or a blocked market opportunity.

That should concentrate minds.

But it should also energise us. Because the companies that get this right will not just comply better. They will compete better. They will build cleaner supply chains, more resilient products, more credible claims, and smarter operating systems.

And that is exactly the kind of climate action we need now.

Practical.
Measurable.
Commercially real.

If you want to understand where this is heading, listen to my Climate Confident conversation with Stephen Jamieson of SAP Sustainability. It is one of those episodes that reframes the issue: carbon data is no longer just about reporting what happened.

It is becoming part of whether you get to sell what comes next.

If this resonates, it makes a strong keynote. I speak on sustainability, AI, and the future of supply chains – details here.