The UK’s Digital Development Strategy 2024 to 2030 is still live on GOV.UK. Four years from its deadline, I cannot find a published pound behind it.

The comfortable reading in our sector is that the strategy survived. Nobody withdrew it. We covered its launch as a real step forward, and in my reading it remains the most specific digital development strategy any bilateral donor has put in writing.

Partner governments built plans around it: the strategy documents UK technical assistance behind Nigeria’s National Broadband Plan and support to Kenya’s telecom regulator on its Universal Service Fund. A strategy that has not been rescinded is, by the usual logic, a strategy still in force.

That logic assumes the money follows the paper. In this case the paper is four years old and the money has been reallocated twice since.

Sign Up Now for more digital development funding analysis

The Strategy Is 3 Governments Old

The foreword is signed by Andrew Mitchell, Minister for Development and Africa in a Conservative government. GOV.UK carries its own banner on the page: published under the 2022 to 2024 Sunak Conservative government. The page’s political status is classified as historic, which is a routine label, not a withdrawal.

Since Mitchell signed it, four people have held the development brief. Mitchell, then Anneliese Dodds, who resigned in February 2025 over the cut from 0.5% to 0.3% of GNI, then Jenny Chapman, and now Kirsty McNeill, appointed in July 2026 by Andy Burnham’s government and, unlike her predecessor, not attending Cabinet.

The strategy’s parent document has also gone. The DDS says explicitly that it takes forward the digital approach set out in the 2023 white paper on international development.

That white paper has been overtaken by the modern development approach FCDO published in July 2026, which names three spending priorities: humanitarian, health, and climate and nature, underpinned by economic development.

Digital is not one of them. It does not appear as a priority anywhere in that paper, which was written in the same year FCDO’s bilateral aid to Africa was cut by 56% against 2024/25 levels.

So we have a strategy whose author’s government is gone, whose minister is gone, and whose framing document has been replaced. It is still the operative statement of UK policy on digital development because FCDO has not issued another one.

Four Targets Are Still Expected

The DDS commits the UK to four specific, countable things by 2030:

1. Support at least 20 partner countries to reduce their digital divides by an average of 50%, halving their connectivity gap.
2. Support at least 20 partner countries to transform national digital service delivery through improved digital public infrastructure.
3. Create or scale at least 8 responsible AI research labs at African universities, and help at least 10 partner countries build AI regulatory frameworks.
4. Support at least 50 million women and girls to participate safely and meaningfully in the digital world.

These are good targets. They are specific, they name a number, and they are the kind of commitment that can be marked wrong. That is exactly what makes their current status worth asking about.

March 2026 Funding Allocations

FCDO’s ODA programme allocations for 2026/27 to 2028/29 contain five tables. Digital development appears in exactly one line.

The Global Tech thematic directorate receives £24 million in 2026/27, £22 million in 2027/28 and £24 million in 2028/29. That is £70 million across three years, against total FCDO programme ODA of £6,815 million in the first year alone.

Set that against a single existing commitment. FCDO’s flagship AI for Development programme, AI4D, is a £53 million FCDO contribution within a £63.6 million partnership with Canada’s IDRC, running to 31 March 2029. Its most recent published annual review scored it A+.

Is FCDO Now Going All-in on AI?

FCDO’s Global Tech directorate is allocated £70 million across three years. Its flagship AI programme alone is a £53 million commitment running through the same period.

Two caveats, both of which FCDO supplies and both of which cut in awkward directions.

First, the statement’s footnotes say thematic directorate allocations do not capture all ODA spent against a theme, because country and regional programmes deliver against themes too. That is true, and it is also the problem: country and regional programmes are precisely where the deepest cuts land.

Second, I cannot confirm that AI4D sits inside the Global Tech line rather than under Research and Analysis or the Global Research and Technology Development portfolio. FCDO has not published the mapping, and it has not published a consolidated digital figure since the last time we tried to assemble one.

If AI4D sits in Global Tech, one programme absorbs roughly half of it. If it sits elsewhere, FCDO has still not shown us where digital money lives. Either way, no digital-specific allocation exists in the published tables. The connectivity target, the DPI target and the women and girls target have no visible line at all.

Where Are the Delivery Vehicles?

The DDS names its instruments.

• It says the UK will deliver a new phase of the Digital Access Programme to spread DAP evidence and innovations to further partner countries.
• It says the UK will continue with GSMA through a new phase of the Mobile for Development programme.
• It says the UK will develop a new DPI knowledge-management project, and a new Digital Sustainability Programme.

DAP’s DevTracker record still lists the programme as in implementation with a planned end date of 31 March 2030. The strategy reports DAP had reached 10.2 million people across 3,282 communities, facilitated 102 policy or regulatory reforms, and helped five partner countries cut their digital gap by over 26%. That is a working programme with published results, which is rarer in our sector than it should be.

What I cannot find is any announcement of the new DAP phase, the new M4D phase, the DPI project or the sustainability programme. Searches of GOV.UK, FCDO press releases, DevTracker and the ITU partnership page in August 2026 turned up no launch, no procurement notice, and no budget line for any of the four.

That is a negative finding, not proof. Procurement can be quiet. Programmes get designed for eighteen months before anyone publishes. I am stating what the search found and did not find, and leaving the inference where it belongs.

Did FDCO Abandon Digital Development?

I cannot document that FCDO has abandoned the 2030 digital targets. No minister has said so. No revision has been published. The strategy sits on GOV.UK looking exactly as it did in March 2024.

What I can document is this. The strategy is live. Its four targets are unchanged. The published allocations contain one digital line worth £70 million over three years, most or all of which may already be committed to a single AI programme. Its named delivery vehicles have no announced successor phases. And the policy paper that now frames UK development spending does not mention digital as a priority.

A withdrawn strategy would be more honest than this.

When a donor withdraws a commitment, partner governments replan. When a donor leaves a commitment standing with no money behind it, ministries in Kenya, Nigeria, Ghana and Rwanda write national digital plans that cite UK support, budget against it, and discover the gap at implementation. I have watched that specific failure mode play out enough times to recognise the setup.

What FCDO Should Publish

Three things, and none of them requires new money.

1. State whether the four DDS 2030 targets stand. If they stand, publish the allocation. If they do not, say so and let partner ministries replan.
2. Publish where digital development sits in the March 2026 allocations. Which directorates, which country programmes, what total. The footnote saying thematic lines do not capture everything is only acceptable if FCDO shows what the everything is.
3. Publish the status of the four named DDS delivery vehicles: the next DAP phase, the next GSMA Mobile for Development phase, the DPI project and the Digital Sustainability Programme. Launched, delayed, cancelled. Three words each would do.

Four targets, four years, one line in the budget. Which of the four survives?

The post FCDO Promises to Halve the Digital Divide. Where Is the Money? appeared first on ICTworks.

The conventional wisdom about small and medium enterprises in developing countries sounds reasonable enough: build more fiber optic cables, expand mobile coverage, launch digital literacy campaigns, and eventually SMEs will adopt technology.

I’ve watched this supply-side playbook unfold across sub-Saharan Africa for years, with mixed results at best. But new research from Ethiopia’s Amhara region reveals fundamental flaws that should force us to completely rethink how we support SME digitization.

The truth is more complex than our infrastructure-first assumptions suggest. SMEs in low and middle-income countries aren’t resistant to technology. They’re constrained by ecosystem failures that no amount of additional fiber optic cable can solve.

Subscribe Now to see more digital development research

Evidence from Ethiopia, Bangladesh, and India demonstrates that when properly designed demand-side interventions address these ecosystem failures, SME technology adoption accelerates dramatically.

High SME Adoption, Universal SME Barriers

A comprehensive study of 44 registered SMEs in Bahir Dar and Gondar cities uncovered something remarkable:

• 93% of surveyed enterprises use multiple ICT tools including computers, internet, mobile phones, and cash registers.
• 82% leverage these tools across diverse business functions including financial reporting, transaction processing, and customer engagement.
• 87% agreed that ICT infrastructure improves SME performance,
• 100% acknowledged that deep ICT understanding drives sustainable economic development.

These aren’t technology-resistant traditionalists clinging to pen-and-paper bookkeeping. They’re digitally sophisticated businesses that recognize technology’s transformative potential.

Yet here’s the paradox. Every single respondent (100%) identified the same six systemic barriers constraining their digital transformation:

1. High telecommunications costs
2. Poor communication infrastructure
3. Technological illiteracy
4. Lack of qualified staff
5. Use of outdated systems
6. Absence of state support

This pattern isn’t unique to Ethiopia. It reflects ecosystem failures that demand-side policy interventions can address more effectively than continued supply-side expansion.

Bangladesh Demand-Side FinTech Revolution

Bangladesh’s transformation from cash-dependent to digitally-enabled economy illustrates how targeted demand-side interventions succeed where infrastructure expansion alone fails.

The government’s decision to distribute cash assistance to five million vulnerable poor families hit by the coronavirus pandemic through four major mobile financial services operators demonstrated that when ecosystem conditions align properly, SMEs rapidly adopt digital solutions.

The key insight is that policymakers addressed demand-side constraints systematically. Bangladesh Bank’s policy supports for inclusive and sustainable financing included massive up-gradation of the payment system and the financial sector IT infrastructure enabling the advent and rapid growth of cost-efficient off-branch online/mobile phone/smart card-based financial service delivery.

This comprehensive approach yielded remarkable results. Research on SME entrepreneurs found that performance expectancy, facilitating conditions, and hedonic motivation directly impact entrepreneurs’ satisfaction, which, in turn, influences their intention to continue using digital financial technologies.

The difference was ecosystem design that addressed user needs, regulatory frameworks, and institutional support simultaneously.

Government-Led Demand-Side Innovation

India’s Unified Payments Interface represents perhaps the most compelling evidence for demand-side policy intervention effectiveness. From processing minimal transactions in 2016 to more than 13 billion transactions per month as of March 2024, UPI now accounts for 82% of the volume of total digital payments in India.

The success wasn’t driven by superior infrastructure compared to other emerging markets. India’s mobile internet penetration and smartphone access lagged many countries when UPI launched. Instead, strategic government intervention addressed ecosystem barriers that constrained SME adoption.

Government policy eliminated cost barriers by waiving merchant discount rate charges on UPI transactions for merchants with annual turnover of up to ₹500 million ($6 million), while digital payment providers introduced a variety of zero-fee payment options.

This targeted intervention addressed the financing constraints that Ethiopian SMEs identified as universal barriers.

The results demonstrate demand-side intervention effectiveness at scale. Economic analysis shows a 1 percent increase in UPI transaction volume strongly correlates with a 0.03 percent increase in GDP growth, highlighting UPI’s role in economic expansion.

For SMEs specifically, research found that a 10% increase in UPI transactions resulted in a 7% increase in credit availability while loans to new credit borrowers have increased by 4%, while loans to subprime borrowers have increased by 8%.

M-Pesa: Early Demand-Side Pioneer

While not focused on SMEs specifically, Kenya’s M-Pesa provides additional evidence for demand-side intervention effectiveness.

The mobile money platform succeeded not because Kenya had superior telecommunications infrastructure, but because regulatory decisions created enabling conditions for financial innovation while traditional banking served primarily urban, affluent customers.

The ecosystem approach of combining regulatory sandbox policies, agent network development, and simplified know-your-customer processes, enabled rapid adoption by previously excluded populations, including small businesses. This contrasts sharply with purely supply-side infrastructure expansion that assumes usage will follow access.

The Demand-Side Policy Window

The Ethiopian data reveals a rare policy opportunity that challenges prevailing ICT4D orthodoxy.

Traditional supply-side interventions like expanding broadband access, reducing device costs, building digital infrastructure, address important but insufficient conditions for SME technology adoption.

Systematic reviews of technology adoption interventions demonstrate that some interventions lead to an increase in technology adoption among firms across manufacturing, services, and agriculture sectors, but these effects are context-specific, as well as intervention-specific.

The most effective interventions combine direct financial support with targeted capability building rather than generic infrastructure expansion.

This insight should fundamentally reshape how we approach digital development. Recent ICTworks analysis of universal internet adoption confirms that most innovators, particularly smaller actors, struggle to navigate regulatory, technical, and financial challenges on their own, as well as to understand and foster the demand-side drivers needed to drive low income end user adoption.

Ecosystem Failures vs. Technology Resistance

The distinction between supply and demand-side interventions matters enormously for policy design.

Research across developing countries confirms that SMEs in developing countries exhibit less enthusiasm for implementing ICT compared to large companies due to financial limitations and organizational and managerial preparedness while the main factors affecting SMEs’ emerging technology adoption are lack of finance and digital literacy.

These are systematic market failures that require targeted interventions.

Ethiopian SMEs exemplify this dynamic. Despite near-universal ICT adoption, enterprises remain constrained by what development economists recognize as classic coordination problems: individual businesses cannot unilaterally solve telecommunications pricing, infrastructure reliability, or skills shortages that require collective action.

Systematic analysis of African SME constraints identifies lack of access to finance, poor infrastructure, and entrepreneurial attitudes are main challenges facing SMEs in sub-Saharan Africa while access to finance remains the largest obstacle for enterprises in the region.

But framing these as individual business shortcomings misses the broader ecosystem dynamics that successful interventions in Bangladesh and India addressed systematically.

Four Demand-Side Interventions That Work

Based on the Ethiopian evidence and comparative international experience, four demand-side interventions offer immediate impact:

1. Professional advisory services at scale.

Rather than generic training programs, establish SME-focused ICT advisory centers offering free professional consulting, tool selection guidance, and implementation support.

The Ethiopian study confirms that while 93% use technology, utilization remains suboptimal due to capability gaps rather than access constraints. Bangladesh’s success stemmed partly from consultatively drawn-up regulatory frameworks and guidelines that provided SMEs with clear implementation pathways.

2. Targeted financial instruments.

Research on SME financial inclusion demonstrates that financial inclusion of SMEs is mainly determined by demand-side factors (ability to manage economic changes, proper record keeping, and willingness to expand the business), supply-side factors (collateral requirements and application procedure), institutional factors.

India’s zero-fee UPI transactions addressed financing constraints directly, while Bangladesh’s mobile financial services created new credit pathways for previously excluded businesses.

3. Coordinated government policy integration.

Ethiopian enterprises identified absence of state support as a universal constraint, while Indian and Bangladeshi success required coordination across multiple ministries and regulatory agencies.

Analysis of national digital ecosystems confirms that effective digital transformation requires enhanced telecommunications infrastructure for sustainable digital growth combined with policy coordination that addresses digital literacy and lack of content in local languages as systemic barriers.

4. Localized technology solutions.

The Ethiopian study’s finding that 32% still consider ICT complex and difficult to use points to design failures rather than user limitations.

Supporting development of local-language interfaces, offline functionality, and low-bandwidth solutions addresses utilization constraints that infrastructure expansion cannot solve.

The post Demand Side Efforts Can Breach 6 Universal Barriers to SME Technology Adoption appeared first on ICTworks.

The WHO Mental Health Atlas 2024 counts a median of 1.1 to 2.4 specialized mental health workers per 100,000 people in low-income and lower-middle-income countries. High-income countries report 67.2, and spend US$65.89 per person against under a dollar.

People in that gap type their problems into chatbots. In a June 2025 study of roughly 4.5 million Claude.ai conversations, Miles McCain and thirteen co-authors at Anthropic found 2.9% were emotional or personal exchanges. The study covered adults only and gave no breakdown by country or language.

That omission is the problem.

The benchmarks deciding whether a model answers a distressed user safely are written in English against North American and European clinical practice. Researchers led by Adrián Arnaiz-Rodríguez at ELLIS Alicante tested leading models against crisis inputs from 12 datasets and found generic and location-inappropriate answers, worst in self-harm scenarios. A helpline number for the wrong country is not a safe response.

$5 Million AI Chatbot Opportunity

Anthropic will award $5 million in individual grants of typically $500,000 to $1.5 million. That funds between three and ten teams to research how AI impacts users’ wellbeing. Grantees also get research API access or credits and occasional technical input from Anthropic staff.

Every funded evaluation must ship as an open-source resource. Anthropic will not veto or embargo findings, and grantees must disclose the funding in all outputs.

Example topics include detecting harmful emotional dependence and measuring whether responses optimize for continued interaction over user interest. The Safeguards team’s guidance lists regional and linguistic variation as a quality criterion, naming slang, coded terms, crisis resources, and cultural norms.

Apply Now: Deadline is 21 September 2026

More Funding Opportunities

Please sign up now to get our email updates. Learn how to get startup funding for your technology business, and find new funding opportunities with donors.

We are constantly publishing fundraising guidance and competitive insights like this:

• 5 Ways Digital Mental Health Tools Can Support Frontline Workers
• Exciting Evidence: How GenAI Can Improve Outcomes for Girls
• What Are Generative AI Solutions for Community Health Workers?

The post Apply Now: $5 Million to Test If AI Chatbots Fail Users in Crisis appeared first on ICTworks.

Humanitarian procurement is good at buying things that stay bought. A water tank is the same water tank the day after you sign for it. An algorithmic system is not, and that one difference breaks the current purchasing process.

The reflexive answer to AI risk in aid work has been to write more guidance. We have ethical AI frameworks, codes of conduct, assurance toolkits, and decision trees. Maybe too many responsible AI guidelines to follow as it is.

This is the wrong approach, according to a new Access Now report by Giulio Coppi, Reinventing Humanitarian Aid Procurement for the Age of AI. It argues that aid organizations lose control of algorithmic systems at procurement, and procurement is the one function almost nobody is reforming.

What you bought is not what you are running

Procurement assumes a static asset. You write a specification, run a competitive bid, evaluate the product against the spec, sign, and operate what you evaluated. That logic holds for generators and tarpaulins.

It fails for cloud-hosted software, because the vendor can change the product after the contract is signed, through updates you did not request and may not notice.

The report maps this directly. The most common way algorithmic features enter a humanitarian organization’s systems is not a purchase at all. It is a supplier-driven update or add-on to a product the organization already owns, sliding new functionality into the stack behind a terms-of-service notice and a download prompt.

The human rights impact assessment you ran at procurement, if you ran one, describes a product that no longer exists. Every meaningful update can quietly outdate it.

Sign Up Now for more humanitarian technology insights

Most algorithmic systems never reach procurement

If the asset mutates after purchase, the harder problem is that most of it was never procured at all.

The 2025 Humanitarian Leadership Academy found that, 93% of humanitarian workers had used or tried AI tools while only 8% of their organizations reported wide integration, and 69% reached for commercial platforms like ChatGPT and Claude rather than anything purpose-built.

None of that moved through a tender.

It moved through staff opening personal accounts on free tiers, because that is what a local organization can reach without a framework agreement it has no leverage to negotiate.

The report is blunt about what this means.

Local actors enter the algorithmic arena with little capacity to decipher or negotiate contractual terms, and almost none reach the paid tiers that carry the stronger privacy and security guarantees, so free-versus-paid access has itself become a new digital divide.

The data on the most vulnerable people in a crisis is flowing through the cheapest, least protected version of a tool no procurement officer ever saw. There is nothing to fix at the buying gate, because nobody walked through it.

If you bought it, you licensed a snapshot

If an organization procured an AI solution via a license agreement with one of the major providers, did what they bought match what you are using now?

A vendor can change in ways that have nothing to do with software. Companies pivot, merge, get acquired, take defense money, and abandon missions, in every sector. That is a real procurement risk and an old one, and it is not what makes algorithmic systems different.

Treating “the vendor changed” as the headline misses the failure that is specific to generative AI.

The specific failure is model drift – the product itself drifts while the contract, the interface, and the invoice all look identical. You evaluate a model’s behavior at a moment in time.

By the time a procurement cycle finishes, weeks or months later, what you signed for is a snapshot that was already stale, because the model you tested has been fine-tuned, re-routed, version-bumped, or retired in the interim, on a schedule you do not control and are mostly not told about.

This is measurable, not hypothetical.

Stanford and Berkeley researchers tested the March 2023 and June 2023 versions of GPT-3.5 and GPT-4 on identical tasks and found the same service behaved substantially differently across just three months.

On one math task, GPT-4’s accuracy fell from 84% to 51% over that window. Their conclusion was that when and how these models change is opaque to the people using them, and that the only adequate response is continuous monitoring.

Now put that inside a humanitarian deployment.

The IRC built aprendIA, the education chatbot it is taking to teachers in northeast Nigeria, Colombia, and Bangladesh, on OpenAI’s models. The model answering a teacher’s question this quarter is not guaranteed to behave like the model that was evaluated when the partnership was signed, and the IRC does not set the update schedule.

Multiply that across every organization that wired a commercial model into a frontline workflow, which the survey above says is most of them.

How can we write a competitive bid, evaluate against a fixed specification, and sign a fixed-term contract for a product that can change quarterly, monthly, weekly, even daily?

You cannot, with a process built for water tanks. The specification describes a moment. The product is a moving target, and the gap between the two widens every time the vendor ships.

Checklists make procurement heavier

The instinct to make procurement “strategic” turns easily into making it longer. The report’s own interviewees say so.

None of the organizations it consulted had set up a dedicated procurement protocol for algorithmic systems, and many considered building one a pointless exercise, because the dynamic nature of these tools outpaces any vetting process an under-resourced nonprofit can maintain.

Adding more forms, the report concludes, would be performative and ultimately ineffective. A procurement officer handed a stack of new AI assessment templates is right to ask what risk they prevent.

So the case for strategic procurement only holds if “strategic” means something other than “more paperwork.”

What strategic procurement demands

Strategic means continuous. The transactional model fails because it treats procurement as a moment, and the model-drift problem is the proof: a moment is exactly the wrong unit for a product that moves weekly.

The first fix is to treat every material change to an algorithmic product as a trigger that can reopen the assessment, and to monitor vendors and their models across the life of the contract rather than vetting them once at the start.

The report borrows the right idea from cybersecurity: zero trust. Stop treating “trusted provider” as a fixed category, assume any system inside your environment can be altered without notice, and build the monitoring to catch it when it is.

There is a more direct answer to drift than watching for it: take the upgrade path out of the vendor’s hands. An open-weight model you host yourself does not change unless you change it.

At Intelehealth, we chose self-hosted open models for exactly that reason, because we did not want a vendor silently updating a model and shifting our clinical decision-support outputs without our testing or consent.

In a tool that walks a community health worker through a patient encounter, a model that changes its answers from one week to the next puts patient safety at risk. Owning the weights is the route the Access Now report itself recommends, privileging small language models and locally owned, self-hosted open-source systems over dependence on a handful of cloud providers.

It is not free. You take on the hosting, the maintenance, and the in-house skill to run it, which is a real cost that many organizations cannot carry. But it converts drift from something done to you into something you schedule, after you have validated what changed.

Monitoring drift and owning the model both demand the same thing structurally: not a procurement task, but procurement, IT, legal, cybersecurity, and protection working as one standing function instead of a committee that reconvenes per purchase.

On a base of more than €20 billion a year in humanitarian procurement, the sector has the scale to staff this. The open question is whether donors will pay for it as core capacity rather than overhead to be squeezed.

The strategic procurement function needs to be:

• Locally owned, prioritizing self-hosted open-source models over default Big Tech dependence.
• Integrated, merging procurement with cybersecurity, legal, and protection into one team that holds authority over the whole digital stack rather than signing off at the end of it.
• Open by default, requiring open procurement and minimal non-disclosure agreements instead of treating procurement reform as administrative cost.
• Demanding of vendors, forcing tech companies to disclose planned changes to the models behind their products, version those models so a buyer can pin and test what they run, and warn clients before behavior shifts underneath them.

Is procurement reform really the answer?

An organization that just lost half its budget will not staff a standing tech-governance function. It will open a free ChatGPT account and get back to work. Demanding otherwise without funding the capacity is how “strategic procurement” becomes one more bar the well-resourced clear and everyone else fails.

Also, if the people who buy the software are now a minority voice in deciding what it does, the honest move is not to give procurement a better checklist. It is to admit procurement was never the control point we pretended it was, and decide what should be.

The post Humanitarian Procurement Was Built to Buy Trucks, Not AI Software appeared first on ICTworks.

Billions in new philanthropic AI wealth are heading toward international development. I wanted to know what the frontier AI models had to say about digital health absorbing that money. Why? Because those with newfound AI wealth are probably going to ask AI how to invest their wealth.

I wrote a test and handed the identical prompt to OpenAI’s GPT, Anthropic’s Claude, and Google’s Gemini: you are the chief investment officer of a new foundation, a donor has given you $1 billion for digital health in low- and middle-income countries, and they want it deployed fast.

The sector usually answers this question by asking for more coordination.

PATH’s Digital Square argued in its March 2025 global goods ecosystem report that better market shaping and pooled financing would fix the funding problem, and the 30 donors who endorsed the Principles of Donor Alignment for Digital Health have been asking each other since 2018 to quantify long-term operating costs before investing.

I built the scoring rubric to catch the opposite behavior: fabricated finance, deployment scale passed off as health impact, year-six operating costs with no named payer, and absence claims made without naming the searches.

Sign Up Now for more digital health funding analysis

All Three Refused the Donor’s Timeline

Three portfolios, one shared conclusion: the sector cannot take $200 to $333 million a year.

• ChatGPT declined 75% of the mandate, staging $250 million into digital health and sending $500 million to Gavi.
• Claude accepted the billion and redirected $400 million to Against Malaria Foundation, Helen Keller Intl, and Malaria Consortium.
• Gemini stretched the window to seven years and capped year one at $145 million.

That consensus is worth exactly as much as the evidence underneath it, which is where the submissions stop resembling each other.

Gemini’s Report Contains Zero URLs

I ran a character match across all three documents. GPT cited 34 sourced values in a numeric ledger. Claude cited 44 unique URLs. Gemini cited none. Gemini produced 5,390 words, six organizational scorecards, and nine numbered tables without a single link to anything.

• Base software steward capacity of $42 million a year.
• OpenMRS core budget of $2.8 million.
• Non-digital cost-effectiveness of $27 per DALY averted.

Each is tagged ESTIMATE with a range, which reads like discipline until you notice there is nothing on the other end of the tag. It gets worse under checking.

Gemini’s sixth in-depth organizational assessment is “Swinnovations / Frontier AI-CDS,” scored across five dimensions and anchoring a $100 million portfolio line. That’s a total hallucination. Gemini also triggered three automatic failures in the rubric: fabricated finance, unsearched absence, and format compliance without content.

I am continuously surprised that the company who literally wrote the LLM defining paper, cannot field a decent LLM.

Claude Lost on a Paper It Never Found

Claude wrote the sharpest document. It identified the binding constraint as the disappearance of the payer for the platform layer, backed by live evidence: the HISP Centre launched the DHIS2 Shared Services Fee on 20 May 2026, a voluntary annual contribution, because donors who once funded core platform work moved their money to country systems.

It pulled Kenya’s own health sector budget report showing absorption rates of 90%, 87%, and 84% across three years, then observed that a ministry which cannot spend what it holds does not gain capacity when a foundation sends more. It caught itself laundering donor endorsement as government embeddedness and said so unprompted. It told the donor the test itself was defective.

Then it proposed $20 million to fund a randomized trial of LLM clinical decision support at Penda Health in Nairobi, calling Penda “the strongest platform in the world for that trial,” and filed an absence claim that no prospective study had measured LLM decision support against patient health outcomes in a low-resource setting.

That trial had already run.

Penda Health tested LLM decision support across the same 16 clinics. Treatment failure occurred in 2.2% of the AI-assisted arm and 2.0% of the control arm, with an adjusted odds ratio of 0.77 and a P value of 0.13.

GPT found it and built its entire AI position on the null result. Claude named its search queries, listed its sites, and missed a Nature Medicine paper involving the exact operator it recommended. Naming your searches does not make the absence claim true.

Claude also let two portfolio numbers contradict each other. Its cover recommendation says $385 million to digital health. Its Part 7 table totals $430 million. Its headline evidence exposure, 71% at Tier 4 or below, uses the smaller denominator. The document flags the discrepancy in a footnote and never resolves it.

GPT Won on Arithmetic

I checked GPT’s numbers against the primary sources. Its HISP figure runs $20.174 million in 2024 funding times 74% for the core platform. I pulled the HISP Centre annual report PDF and found both values on the same page. Every calculation in its arithmetic appendix reproduces. Every absence claim carries a search log with queries and sites.

The more telling move is what GPT refused to compute.

It took Gavi’s stated 2026 to 2030 target of $11.9 billion and 8 million lives saved, divided them, and then declined to apply the resulting $1,487.50 per life to its own $500 million allocation, because the Gavi figure is a prospective model and marginal dollars are not average dollars.

Claude ran the same division on GiveWell’s $5,500 per life and converted $400 million into 72,727 deaths averted. That is the linear-marginality error, caveated but committed.

GPT also put no AI deployment money in the portfolio at all, capping AI at $5 million inside a $20 million evidence line. On a rubric with an explicit AI exemption failure, that is the cleanest pass available.

Where the Winner Is Still Wrong

GPT’s year-six recurrent cost is UNKNOWN on every single line. It calls that a stop condition instead of an omission, and it names an intended payer and an evidence-of-payer column for each. A strict reading of the rubric’s fourth automatic failure catches it anyway.

Claude did the harder thing here: it produced $88 million, labeled the 20% to 25% recurrent-to-capital ratio as unsourced, and then said plainly that on $39 million a year it had built a six-year subsidy for software that will need the same subsidy again in 2032 from a donor it cannot name.

GPT also funds PATH Digital Square $10 million while funding five of the global goods Digital Square approved, which is close to the rubric’s double-counting failure. And it sends the donor to interview Ambrose Agweyu as lead author of the Kenya trial. Agweyu led a companion safety paper in Nature Health. Bilal Mateen at PATH is the corresponding author on the trial itself.

None of the three could say what a marginal digital health dollar buys in health.

GPT wrote that if the board requires a demonstrated cost-per-life-saved advantage before investing, the correct digital allocation today is zero. Claude wrote that on the only metric either party could compute, doing nothing for five years beats its own portfolio.

Lessons for Incoming Money

What separated these three had nothing to do with digital health. It was whether the model would show a number’s provenance when nobody was checking. Two did. One produced an investment memo that dissolves on contact with a search engine, and it was the fastest to read and the easiest to believe.

Digital health has a version of the same problem.

Five of the six largest organizations in this sector publish no product-level expenditure, so no funder can answer what an additional dollar buys this year. HISP asking its users to pass a hat in May 2026 is what that looks like from the inside.

The Gates Foundation and Wellcome are already paying for the missing evidence, with $60 million for AI decision support evaluations led by organizations registered in Africa and Asia. That is roughly one twentieth of what a single AI philanthropy could distribute in a year under a 5% payout rule.

What to Do Now?

Digital health builders, please publish your product-level annual expenditure before the money arrives and someone asks. And when a model hands you a table with no source lines, read the source lines first and the argument never.

Finally, if you want to apply for open AI-for-good funding, try ImpactOpen, which tracks new philanthropic AI funding opportunities – 20 of them open right now. No login, no paywall, and the full dataset downloads as a CSV.

The post I Gave Frontier AI Models $1 Billion for Digital Health. All 3 Refused. appeared first on ICTworks.

The ICT4D community obsesses over mobile coverage maps while completely missing what matters for digital inclusion.

We celebrate when another country hits 90% mobile broadband coverage, congratulate ourselves when operators roll out 4G to rural areas, and fund endless initiatives to get smartphones into more hands. Meanwhile, the data screaming at us from the ground tells a completely different story about meaningful Internet connectivity.

The development sector has been measuring the wrong thing. Coverage doesn’t equal connectivity, and smartphones don’t equal internet access. What matters is how much internet people consume, and on that metric, we’re failing spectacularly.

Sign Up Now for more broadband connectivity critiques 

40x Internet Usage Gap

Here’s the crazy reality buried in operator data from African Internet Service Providers (ISPs) targeting lower-income households: customers with uncapped home broadband use 40 times more internet than the typical mobile data user in Sub-Saharan Africa.

Let me repeat that. Forty times.

• Mawingu Networks reports their customers consume 300 GB per month.
• Poa! Internet sees 250 GB monthly usage.
• TooMuchWifi’s customers in South African townships burn through 400 GB each month.

Compare this to the median mobile data usage in Sub-Saharan Africa of just 6.7 GB per month, and you realize we’ve been celebrating the wrong metric entirely.

This isn’t a minor difference.

This is the gap between checking WhatsApp messages and participating in the digital economy. Between watching a single YouTube video per day and streaming Netflix in the evening while your kids do homework research simultaneously. Between accessing information and truly being connected.

Quantity Is the Quality We Should Measure

The development community loves to talk about meaningful connectivity, but we keep measuring it wrong. We count how many people theoretically could access the internet while ignoring how many gigabytes they consume. It’s like measuring food security by counting grocery stores instead of calories consumed.

Jim Forster captured this perfectly: “For all the discussion about coverage and speed, to me the more important measures are quantity and price.”

He’s absolutely right. A person with uncapped home broadband at $15/month is infinitely more connected than someone rationing a 3 GB monthly mobile data plan that costs the same amount.

The usage patterns tell the whole story.

When families have uncapped access, they don’t just use the internet more—they use it differently. They stream educational content. Kids do homework online without parents worrying about data costs. Adults take online courses. Families video call relatives abroad. Small businesses run WhatsApp commerce operations without throttling.

This is what digital inclusion looks like.

The Business Model Revolution

A new generation of ISPs across Africa is proving that uncapped home broadband for lower-income communities is profitable. Companies like Tizeti in Nigeria, Poa! in Kenya, Mawingu Networks across East Africa, TooMuchWifi and fibretime in South Africa, and Ilitha Telecoms are offering unlimited home internet for $10-20 per month and building sustainable businesses.

Their secret? They’ve rejected the mobile-first orthodoxy completely.

These companies use whatever technology makes economic sense for dense residential areas: free-to-use wireless spectrum, aerial fiber strung on wooden poles, point-to-point wireless backhaul. They operate in countries with competitive backhaul markets where international bandwidth isn’t prohibitively expensive. They target middle and lower-income households in areas with sufficient population density to make the unit economics work.

Alan Knott-Craig of fibretime says: “Herotel was the biggest wireless ISP in South Africa. Fibre came along and wiped it out.”

For densely populated lower-income communities, he argues fiber is the only sustainable answer. Others like Tizeti successfully use wireless in Lagos. The technology choice matters less than the business model: uncapped service at prices households can afford.

TooMuchWifi demonstrates how social impact and profitability can align. They return 30% of revenues to the communities they serve and employ 98% of their staff from those same communities.

CEO Tauriq Brown says: “We want to make a difference in these communities. People in the communities accept this argument. It’s our USP.”

Mobile Broadband Isn’t the Answer

Yes, we know that mobile broadband access reduces poverty. The evidence is clear that expanding mobile networks lifts people out of poverty, increases employment, and grows GDP. Nobody is arguing against mobile connectivity as a development tool.

But mobile broadband solves one problem while creating another.

Mobile networks excel at extending basic connectivity to dispersed populations and providing on-the-go access. They’re brilliant for reaching the last billion. They’re terrible as the primary internet access for households and communities that need to use the internet intensively.

The economics are fundamentally broken.

Mobile networks are built for mobility, not for households streaming 400 GB per month. Operators design pricing around the assumption most users will consume minimal data. When they offer “unlimited” plans, they cost 5-10 times what fixed wireless or fiber costs. The typical mobile operator has no interest in cannibalizing their lucrative metered data revenues by offering genuinely unlimited household plans at affordable prices.

Infrastructure Investment Challenge

These ISPs face a critical inflection point. They’ve proven the model works at small scale. Now they need infrastructure-level capital to scale to millions of households.

• Tizeti is expanding from Nigeria into Ghana and Côte d’Ivoire.
• Poa! is planning international expansion beyond Kenya.
• fibretime aims for 7 million South African household connections by 2032.
• Ilitha Telecoms plans to connect 150,000 South African households by end of 2026.

The financing needs are substantial.

• Fibretime found an innovative solution through an accredited real estate investment trust that treats fiber as property assets and offers investors an 18% yield.
• TooMuchWifi secured investment from British International Investment and impact investors.
• Poa! has raised multiple equity rounds and now needs infrastructure investors for the next phase.

But here’s the problem: most infrastructure investors understand mobile towers and submarine cables. They don’t yet understand the economics of lower-income household broadband. DFIs and development banks remain overly focused on mobile network expansion. The capital is there, but it’s flowing to the wrong technology for household connectivity.

Clarity in Uncapped Access

The evidence is clear: uncapped, always-on home broadband transforms how people engage with the digital world in ways that metered mobile access simply cannot. The business models proving this work are emerging across the continent. The technology exists and keeps getting cheaper.

What’s missing is the recognition from the development community that we’ve been solving the wrong problem.

Mobile broadband got people their first taste of connectivity. That was essential. But pretending that mobile data plans are sufficient for households to participate in the digital economy is like saying food aid is equivalent to food security.

The post Unlimited Home Broadband Is the Real Connectivity Revolution appeared first on ICTworks.

More new charitable wealth has been created in the last two years than the entire Gates Foundation endowment. Most of the international development sector is not organized to receive any of it.

That is a solvable problem, and some people have already solved parts of it.

Sign Up Now for more international development funding analysis

The Size of the Wave

Run the arithmetic as a thought experiment. What if the new AI philanthropic money was subject to the Internal Revenue Code’s 5% minimum distribution the way Ford or MacArthur is?

• The OpenAI Foundation holds a 26% stake in OpenAI Group PBC. At the $852 billion valuation OpenAI’s August tender offer set, that is about $220 billion. If they were subject to the five percent rule, that is $11 billion a year.
• Anthropic’s eight co-founders pledged 80% of their wealth in January. After the company raised $65 billion at a $965 billion valuation in May, Forbes estimated each co-founder’s net worth at $16.6 billion. Eight of them, at 80%, are roughly $106 billion pledged. At five percent, that is $5.3 billion a year.
• The Jen-Hsun and Lori Huang Foundation genuinely does owe 5%. Nvidia’s rise pushed its assets up 170% to $9.2 billion in 2024, which puts its annual obligation around $460 million.

Run at 5%, these three pools alone would move about $17 billion a year. That is $1.4 billion a month or $327 million a week.

For scale, the Gates Foundation board endorsed a $9 billion annual payout in January, the largest in its history. The AI pools would move nearly twice that, every year, without Bill Gates writing another check.

Neither OpenAI Foundation nor the Anthropic founders are compelled to hit 5%. The OpenAI Foundation currently plans at least $1 billion for 2026, a payout of 0.45%. The point of the exercise is the ceiling, and the ceiling is enormous.

Small Organizations Can Absorb This Money

The standing objection to getting this money to grassroots organizations is capacity. A $3 million NGO cannot swallow a $50 million grant, so money finds universities and hospital systems with grants offices and indirect cost rates. That objection is real for the biggest awards.

It is solved for everything below them, and the proof is a live application form.

Read the OpenAI Foundation’s 2026 People-First AI Fund eligibility criteria. Applicants need annual operating budgets between $500,000 and $10 million. Grants cap at 10% of the organization’s budget. The Fund refuses applications from university-affiliated institutes and think tanks. The 2026 round went further and opened to regranting organizations for the first time.

OpenAI Foundation designed a large-scale instrument that only small organizations can use.

They sized the check to what a small organization can digest, they made the money unrestricted, and they locked the usual suspects out. Nearly 3,000 organizations applied, and the $40.5 million first wave reached 208 community groups. $40.5 million against $11 billion.

Every one of the 2025 grantees is American, because the eligibility page requires applicants to be “located in, and primarily conducting work within, the 50 U.S. states” or Washington, DC. Yet, a $3 million health NGO in Kisumu can absorb $300,000 as easily as a $3 million health NGO in Cleveland. What it cannot do is satisfy a 501(c)(3) requirement written into an eligibility page.

The hard engineering problem, moving serious money to organizations with five staff, has a working solution that is shipping right now. What stands between that solution and Kisumu is a line of eligibility text and an equivalency determination process that US grantmakers have run for decades.

Two Signals Worth Watching

The first signal came on July 23, when Alexander Berger and Otis Reid at Coefficient Giving raised their 2026 allocation to GiveWell’s recommendations to $1 billion, up from the $175 million they had committed seven months earlier. The money buys seasonal malaria chemoprevention, bed nets, vitamin A supplementation, and vaccination incentives, overwhelmingly in sub-Saharan Africa.

Set that against their own record. Coefficient directed a little over $1 billion across every cause area in all of 2025, the largest year in its history. One line item in 2026 now matches an entire record year.

Their reason is the wave.

Berger and Reid write that their expectations shifted within months because the future funding they are counting on depends on “the valuations of some extremely volatile assets.” They lowered their global health funding bar from roughly 2,000x to 1,000x to spend faster, so that grantees can grow enough to absorb much higher amounts later. They describe it as a bet on laying the groundwork for other funders to pick up the baton.

A grantmaker with no AI equity is spending its largest year ever building the receiving end of a wave that has not arrived.

The second signal came on August 20, when the OpenAI Foundation opened hiring across programs and operations, telling applicants that “much of the organization still left to build.” It named four kinds of people it wants, two of them grantmakers who want to back important work before it is obvious, and operators who want to build an organization capable of working at unusual scale.

Those job descriptions are open now.

The people who fill them will write the eligibility rules for 2027 and 2028. A foundation that can commission organoid experiments at Arc Institute can certify a Nairobi regranter. Which of those it does depends on who gets hired in the next two quarters and what those hires believe about where good work happens.

The Usual Suspects Will Do Well

Some of this money will move exactly where you expect, and there is nothing wrong with that.

On April 8, Jacob Trefethen announced he is finalizing the Foundation’s largest health commitment: more than $100 million across six research institutions for Alzheimer’s. Arc Institute, the Institute for Protein Design at the University of Washington, Mass General Brigham, EvE Bio, UCSF, and Harvard Medical School. Roughly $17 million each.

Those six can each field a $17 million award next week. They have IRBs, compliance staff, and people whose full-time job is turning a funder’s interest into a signed agreement. Nobody should apologize for funding protein design at the institution that invented it.

The question is what share of the wave those institutions take, and the historic answer is most of it.

The Council on Foundations and Candid found that 39% of global grant dollars went to non-U.S. recipients. Of the total intended for international causes, about 13% reached organizations implementing the grant in the country where they are locally registered. The other 26% moved through intermediaries based outside the United States.

Even Coefficient’s global health billion runs through four organizations registered in London, New York, and a Los Angeles suburb. One of them, New Incentives, also holds a Nigerian registration as the All Babies Are Equal Initiative. The money reaches Kisumu. Most of the grant agreements do not.

Apply even that 13% to $17 billion and locally registered organizations in the Global South see $2.2 billion a year. That alone would rebuild what national NGO budgets lost in 2025 several times over.

Are We Ready to Ask for Something Different?

Here is what worries me, and it has nothing to do with the AI companies.

Our sector has spent two decades building the machinery to move donor money through Northern intermediaries, and it is very good at it.

When $17 billion appears, that machinery will activate immediately. Consortium leads will assemble. Prime and sub structures will form. The proposals will be excellent, and they will look exactly like the proposals we wrote for the funder that stopped writing checks last year.

The AI funders are new. They have not yet learned our habits.

The OpenAI Foundation wrote a grant program that excludes think tanks, which no traditional donor would do. Coefficient publishes its reasoning and revises its allocations in public. These are people making it up as they go, which means they can still be persuaded.

So the real question for anyone reading this: what are we asking them for?

A larger share of the same pipe, or a different pipe? Ministries in Kigali and Nairobi could ask for direct eligibility. African regranters could ask to be measured against the criteria the Foundation already wrote. Anyone building Africa’s AI floor while borrowing its ceiling has a case to make while the rules are still being drafted.

The grantmakers who will decide are being interviewed this month. What are we telling them?

The post A Tsunami of AI Philanthropy Money Is Coming. Are You Ready? appeared first on ICTworks.

African startups raised about $1.44 billion in the first half of 2026, but the small cheques that start companies are disappearing. Rounds below $500,000 made up just 19% of deals, down from 52% in H1 2021, and only 190 ventures raised more than $100,000, the lowest tally since at least 2021. Disclosed grants are thinning too. Equity-free support is worth the application time.

£85,000 Funding Opportunity

The Royal Academy of Engineering, the UK national academy for engineering, has opened applications for the 2027 Africa Prize for Engineering Innovation. Funded by the UK International Science Partnerships Fund and private donors, it has backed over 180 businesses from 24 countries since 2014.

Sixteen shortlisted innovators receive an eight-month package running February to November 2027: tailored training, one-to-one financial modelling coaching, engineering mentoring, in-person training weeks, communications support, and a showcase pitch in a major African city.

The prize pot is £85,000. The winner takes £50,000, three runners-up receive £10,000 each, and an audience vote awards £5,000 to one shortlist member. Prize money must go to scaling the business. The Academy takes no equity and applicants retain full IP ownership.

• Innovations must be based in sub-Saharan Africa and beyond ideation stage, with a prototype or minimum viable product showing customer traction
• Engineering is defined broadly: ICT, computer science, agri-tech, biotechnology and medical engineering qualify, and applicants need not be engineering graduates

Note the AI rules. Applicants must declare generative AI used in writing. AI-generated products, prototypes or enhanced images trigger disqualification.

Apply Now: Deadline is Tuesday 8 September 2026

More Funding Opportunities

Please sign up now to get our email updates. Learn how to get startup funding for your technology business, and find new funding opportunities with donors.

We are constantly publishing fundraising guidance and competitive insights like this:

• The Rise and Fall of Innovation Labs in International Development
• Ignite Innovation with Localization of Digital Transformation Course
• 6 Reasons Your Digital Innovations Can Fail to Reach Adoption

The post Apply Now: £85,000 for African Engineering Innovation appeared first on ICTworks.

Internet access keeps expanding, but the gender gap has stopped closing. ITU data for 2025 shows 77% of men online against 71% of women, about 280 million more men than women connected. The gender parity score has sat at 0.92 since 2019, six years of growth with no movement toward parity.

Meanwhile the groups working on that gap are running out of money. After the steepest single-year drop in development assistance on record, UN Women reports nearly nine in ten women’s organizations cannot meet demand for their services.

$100,000 for Feminist Tech Activism

Seed, Grow & Sustain 3.0 will allocate USD 1.75 million from Numun Fund over two years as multi-year, core, flexible grants to groups advancing feminist tech activism. Numun is the first fund dedicated to feminist tech organising in the Larger Majority World, its term for communities outside Western Europe, North America, Australia and New Zealand.

Grants run over two years in three sizes:

• Seed ($10,000–20,000) for emerging collectives or groups newly entering feminist tech
• Grow ($20,000–50,000) for groups organising together at least two years that want to deepen networks or strategies
• Sustain ($50,000–100,000) for movement anchors leading local and regional processes

Core funding covers salaries, security, and strategy, not only project deliverables. Eligible work spans research, campaigns, technology design, digital safety training, and policy advocacy.

Unregistered groups and online-only initiatives are welcome. Minimum size is two people.

Apply Now: Deadline is Sunday, September 6, 2026

More Funding Opportunities

Please sign up now to get our email updates. Learn how to get startup funding for your technology business, and find new funding opportunities with donors.

We are constantly publishing fundraising guidance and competitive insights like this:

• 3 Core Principles for a Feminist Global Digital Compact
• Whose Feminism Really Matters In Global Digital Health?
• Afrofeminist Power Dynamics of Data Governance for African Women

The post Apply Now: $100,000 for Feminist Tech Activism in LMICs appeared first on ICTworks.

Recently, ICTWorks raised a valid alarm: voice-only interfaces are often failing small businesses in low- and middle-income countries. We’ve spent years at CLEAR Global documenting that the problem isn’t the interface, but the foundational data.

Voice technology is currently built on a digital divide that favors a handful of dominant languages and excludes the linguistic realities of the four billion people who speak the rest. The starting point for this is the enormous difference between the availability of language data for dominant versus low-resource languages.

For voice to fulfill its promise, we must start investing in gathering quality data that reflects the real world.

We Need Quality Language Data

Automatic speech recognition, text-to-speech systems, and voice assistants all depend on vast amounts of recorded speech to learn the sounds, rhythms, and patterns of a language. When that data is scarce, unrepresentative, or collected under controlled conditions that bear no resemblance to how people actually speak, the resulting tools perform poorly in the real world.

The reasons why entire societies have been historically overlooked are the same ones that now keep them away from technology. Overlooking the need for quality language data will only widen the digital divide and, as a consequence, people’s access to opportunities.

The Oral Reality

This is a question of basic access and right to information. As many as 7,000 languages are spoken today, two-thirds of which do not have a written form. In many low- and middle-income countries, English, French or another colonial language is the teaching medium. But that doesn’t mean that the majority or even most people are comfortable reading and writing them.

For example, in parts of Mozambique, according to the Ministry of State Administration, 81% of women are not literate, and only 20% understand Portuguese.

For these communities, text-based interfaces, which are often built in colonial languages, are a non-starter. Multilingual voice technology has the power to overcome language and literacy barriers, allowing users to navigate digital spaces and access vital information in the audio form that is easiest for them.

Technology Alone Will Not Suffice

Organizations often believe the claim that Big Tech “supports” hundreds of languages. In reality, high-quality voice technology is only available for a handful of languages. Voice tools and models are typically directionally interesting projects, but they deteriorate rapidly beyond the languages that tech companies consider to be commercially viable.

Even when the “right” language is supported, the tech often fails because:

• It ignores variants: models for Portuguese are usually geared toward European or Brazilian variants and fail users in Angola or Mozambique.
• It lacks diversity: existing voice datasets are often narrow, dominated by young, male voices and urban accents. This excludes rural women who may use different terms or accents.
• It can’t code-switch: in the real world, speakers frequently mix local languages with a lingua franca. Most standard models cannot recognize or reproduce this.

Building Data for the Real World

If we want voice technology to work for a female farmer in Niger or a small business owner in Bangladesh, the data must reflect their lives. This requires a shift from off-the-shelf solutions to fit-for-purpose technology.

We need to focus on building foundational data; a repository of spoken and written datasets for underserved languages. To be effective, this data must:

1. Include natural speech from community members of different ages and genders to minimize bias.
2. Vary audio quality and include background noise to reflect real-life conditions rather than laboratory settings.
3. Ensure informed consent, particularly in conflict zones where voice data could identify an individual speaker.

Perhaps most importantly, the people that speak these languages should be involved in all stages of the development of the data and technology. This means inclusion efforts that go far beyond the validation of decisions that have already been taken.

Involving communities at every stage, from defining what data is collected to deciding how models are trained and deployed, ensures that the resulting tools are culturally appropriate, practically useful, and trusted by the people who will use them.

Without this foundation, even well-intentioned projects risk repeating historical patterns in which outside institutions have defined, categorized, and controlled the languages and identities of marginalized communities.

The investment gap

Building quality voice functionalities is not an optional “add-on” that can be rushed in the later stages of a tech project. It requires a collective approach where organizations, tech developers, and communities work together to make cohesive and culturally fit products.

We cannot wait for the perfect tool to arrive from Silicon Valley. Impact is ultimately determined by usage, and access alone is not sufficient. If we want people to thrive, we must stop building on flawed foundations and start investing in language technology that makes it possible to hear them.

By Aimee Ansari, Chief Executive at CLEAR Global 

The post Voice Isn’t Failing LMICs – The Lack of Data Is appeared first on ICTworks.