We’ve seen this before. Centralised typing pools disappeared and became distributed admin assistant roles: a floor of typists became individual admin assistants spread over the building using wordprocessing on personal computers! Telex operators became IT departments. Each wave of technology compresses certain functions and expands others. AI is doing the same thing now: faster, and with more impact on how careers actually progress. The career ladder changes.

What’s different this time is where the pressure hits home. It’s not just entry-level work being automated. Entire layers of management – coordination, reporting, scheduling – are being compressed, while new roles emerge around AI handling, oversight, and leadership judgement. The traditional career ladder is being reshaped from underneath us.

This is NOT just a story about job loss. It’s a story about role redesign. About moving from doing tasks to owning outcomes. About understanding where human judgement still matters: and I encourage you to step into it deliberately.

In this video, I walk through how organisations are restructuring in response to AI, what’s actually changing in the career ladder, and how you can prepare without needing to become a technical “AI Vibe Coder blah blah” specialist. The goal isn’t to keep up with AI; it’s to stay relevant in a system that is reorganising itself around it. I truly hope this helps.

TLDR -what restructure due to AI looks like:

• AI agents doing frontline work
• Human “AI handlers” managing and auditing systems
• A compressed management layer
• Expanded leadership roles focused on judgement and accountability

Laurel Papworth on how organisations restructure around AI and how to manage that expansion and compression.

Transcript of AI Proof Your Career – Career Ladder and Organisation Restructure due to Artificial Intelligence

[00:00:03.24]
Hello, my name’s Laurel Papworth. It’s pretty late in the afternoon, early evening here actually. Um, but I, so my voice is a bit croaky, but I want to get this video out. It’s another one in the series of how to AI-proof your career. And this specific one is about how the organisation’s going to restructure. And the career ladder changes and how you are to prepare for that. You’ve got a little while, but you should know this is going on. We all love a restructure, don’t we? Um, no, we don’t. So it’s not just for employees.I would encourage anybody in HR to consider these key points and maybe also change management consultants, people like that who are affecting change within an organisation, restructuring and things like that.

[00:00:55.10]
So the first thing to recognise, I’m going to talk about what the issues are and then how to overcome them. With emergent tech, major restructures happened. So think about the typing pool that was the whole floor of the office was typists using typewriters, and then they reskilled and upskilled to use personal computers and then word processing packages like, um, WordPerfect and Microsoft Word, and then they were deployed in a in a distributed way. So not a central typing pool, but distributed throughout the whole organisation. Every floor had a personal assistant, an executive assistant, administrative officer.And in some cases there’d be one computer in that department at the beginning, and it was on his or her desk. And they were the people who entered everything into the computer. So you can see the whole organisation restructured. Instead of tonnes, dozens and dozens of typists, These people were redeployed into every office and eventually everybody was reskilled and upskilled. Telex people, the ones, the one or two people that looked after the telex machines, that was how interoffice communication took place. That got shifted into an IT department with mail servers. You get the idea. Every emergent tech has this change that occurs. And I call it compression and expansion. So the typing pool compressed, but what expanded was reskilled people placed within departments. In-house librarians that would find physical documents for people were replaced by databases, database engineers, internet searches, things like that. And you have this compression, expansion, compression, expansion. So when you know that, it’s a good idea to consider how much of your work is going to be compressed, how much would be expanded. In the last video, I talked about taking task bundles, breaking them down into tasks, and then putting them into columns. This one’s AI ready and this one needs human non-negotiable so that you can actually see which tasks you have there. The layoffs have been pretty horrendous.I know there’s AI washing where companies are saying we are laying off staff because of AI, but it’s actually because of poor management of resources and they’re blaming AI. Poor old AI. Who wants to blame AI? But yeah, AI is causing some of it too. So I’m talking here Intuit that make QuickBooks and Mailchimp. You’ve got Cisco Systems. Shout out to Cisco. You’ve got Oracle, IBM. Oh yeah, Dell. And Amazon— so Amazon laid off or is laying off 14,000 corporate office staff. And I specifically want to talk about the type of staff that are being laid off and then the other type of staff that are being hired in place of them. And it’s not technical staff, so I’ll come to that in a moment. You don’t have to go out and get an AI degree. That’s not what this is about. So the compression that’s mostly occurring is in the structural career ladder. It’s the structure of the department, and I just need to kind of explain how this works. So let me take a customer service department.

[00:04:22.28]
You have a customer service agent, human being, who takes the calls and responds. They have a knowledge base, they’re trained, and they answer the questions. If a customer is unhappy, and asks to speak to a manager, then it gets escalated to the team leader or the supervisor. The team leader or the supervisor works for kind of like a department manager, and that department manager works within the system and has like maybe 12 dashboards. There’s a staff scheduling dashboard and there’s a resource allocation dashboard and there’s a procurement dashboard and there’s a request toilet paper dashboard. I don’t know. There’s all kinds of dashboards and they know exactly which system to go into in order to look things up and do what they need to do. They do resource allocation stuff, scheduling, and this dashboard stuff. We’ll come to that in a moment. Above them would be, I don’t know, an area manager and above them would be a chief customer service officer who sits on the executive and they’re in the leadership group. They, the leadership doesn’t work in the system, they work on the system. They initiate changes. Managers don’t change the system. They work with the changes as they come down the line. So that’s been the career ladder. If you’re really good at customer service and then you show yourself to be a real people person, you might get promoted into supervisor because you go above and beyond. From supervisor, you get promoted into department, uh, into the group head and then into, you go off and do some post-grad graduate work and you end up being the one of the chief executives in this area. Awesome. That’s the traditional career ladder.

[00:06:09.05]
As the organisation goes through a restructure, the agents that speak to the customer will be AI agents. So AI agents will be first port of call for customers. They’ll be given something called a decision hierarchy.In human terms, it’s an escalation chart. There’s a few differences, but just call it a decision hierarchy, which says if the customer says they’re going to the ombudsman, if the customer says they’re going to the press, if the customer says my brother’s a lawyer, whatever it is, then the agent and the customer service rep both know to escalate this. And in this case, there’s going to be an influx of AI handlers who are like micromanagers for a team of agents. So if you have core skills and you you know the domain knowledge, you know the customer service knowledge or the technical support knowledge or the whatever it is that you’re doing, you’ll be put in charge of the agents so that they can come to you, your mummy or daddy, whatever, however you identify. But they’ll come and ask you questions. Should I respond to this? Because this is going to be written into the agents. You’ve got to put constraints in which says don’t email back automatically, escalate it if you see this problem.

[00:07:34.17]
Once it goes through to this mid-level management, at the moment, the hiring around AI handlers is primarily interns. So maybe you did a graphic design course and you are straight out of graphic design college. You’ll be given charge of the AI agents that are creating the graphic designs for the company, and your job is to make sure that the images don’t have two heads and 10 fingers and are doing something inappropriate. Inappropriate that could walk you into a brand reputation issue or, um, something catastrophic. It’s problematic because it actually means that graphic designers won’t get skills in doing graphic design. They’ll only be given skills in managing AI agents that are doing graphic design.And I think that’s a concern. A replacement there for customer service where it becomes AI agents, and then there’s an AI handler who is somebody who has domain knowledge or specific knowledge, and then just enough information to be able to take on their calculation, decide what to do with it. Human judgement steps in where there’s accountability. So wherever there’s risk or accountability, put a human in. The management layer vanishes altogether. This is the compression. So you’ve got compression at the agent level, expansion at the handler supervisor level, and then the manager level with all the dashboards. The AI Claude can create a dashboard in seconds. Resource allocation, AI can do that. Staff scheduling, AI can do that. So these manager layers are problematic because they build in efficiencies within a system that an AI can be more efficient at. I’m not saying there won’t be an AI handler at a management level, but there’ll be many, many, many less. So there’s really not going to be many. We’ll get on how to fix this in a moment.

[00:09:28.28]
The next level up, which is, um, into the leadership space, that expands and that expands automatically, because what’ll happen is everything that’s outside of the standard operating procedure will be sent through to a human who takes agency. So if you’re the sort of person that sits at work and goes, nah, above my pay grade, oh, I’m not getting involved in that discussion, that’s worth more than my job is worth, anything like that, you have a problem, Houston. But if you’re the sort of person who likes to jump in and resolve those things and find solutions, then you move into what I would call the leadership layer. And that, that area expands, which is one of the reasons why you are seeing OpenAI and Anthropic and others hiring a lot of content creators that have deep experience in delivering great messages because the AI can churn out low-grade stuff, but they need people with real leadership who are willing to put skin in the game to step up. And drive certain areas outside of the system. So you’ve got AI agents replacing customer service, but then you have an intern level coming in as the supervisor. You’ve got the compression of the management layer, and then you’ve got expansion of the leadership area.

[00:10:54.19]
What are you going to do? Because your normal career ladder is to take each one of those, and now they’re— some of those steps are disappearing. And that’s what happens with AI. It compresses the whole organisation, takes 10 steps person makes it through. I would suggest that you work on specific skills and you don’t worry too much about others.

[00:11:17.18]
Let me explain. First of all, be really honest with yourself and say where you actually sit within the three tiers of an agent, intern supervisor, and leadership. Are you an agent just doing day-to-day work? Are you the supervisor who tries to problem solve, or are you leadership where you take responsibility and accountability? Which one are you most comfortable with? If you are going to stay in the handler area, I suggest you deepen your domain knowledge. So when I worked in, um, troubleshooting, like technical support, there were certain people that could answer lots and lots of calls and resolve them very quickly. I was always the person who went and chased up the answer later on with the vendors or with subject matter experts because I hated not knowing the next time somebody rang what the answer was so that I could tell them. And that paid off very well for me later on because I would be called into very senior meetings because I was the person who understood more about how these pieces went together. So people who quickly churn out answers, that’s going to be a problem for them. AI can do that. But the outliers, the questions that people don’t have a quick answer to, the ones that make you say, oh, lemme just stop and think about that for a moment. They’re the ones that you want to deepen your knowledge on. So that means going very, very deep on that.

[00:12:53.23]
If you want the leadership layer, you have to understand accountability and be willing to take accountability. And this is what accountability looks like. Yep, I’ve listened to everybody. I I think you’ve all got great points. We’re going to have to make a decision here, and my decision is we’re going to pursue this path. And if that doesn’t work out, it’s on my head, not on your head. I will back you and protect you in this. Are we all in agreement? Okay, Jim, you’re never in agreement. Everybody else is awesome. Let’s go. So it’s about taking accountability. It’s about making a decision under pressure. It’s about not letting the AI— the— well, the system said not to do that, like That’s not leadership. You need to take back the reins. You’re the pilot. AI is the co-pilot. Make sure you pilot this ship. So if you feel like you can do that, then awesome. But otherwise you might want to stay a handler. I think middle management, there’s no chance. All those tasks are compressed. They’ll be farmed out. They’ll become— the executive will run their own dashboard and press a button to get the report that used to take a week to write. That’s, I think that’s a no-brainer. So I wouldn’t be looking for a manager within the system. I’d be looking for leadership to change the system or handling very specific, very deep knowledge on specific topics. Okay, so another area to look at would be seam knowledge. I would call this the branches and it’s systems integration. Not just knowing your department, but knowing how your department impacts operations. IT, finance, HR, knowing how all the pieces come together. AI is really good at deep, one subject domain, but it’s not so brilliant across the board. I would recommend developing seam knowledge. Understand how your department integrates with other departments, like really become a business person here. The person who knows how marketing breaks operations and impacts IT is invaluable. It’s not parameters at the moment that AI can manage as well as a human with deep knowledge can. So single domain experts are fully exposed because I can train an AI. A domain just means subject matter. So if you have, I don’t know, deep knowledge on billing and fraud systems, I can chuck all the manuals and all the trouble tickets for the last 10 years and everything at an AI and go, learn this, and it will learn it. So that kind of deep knowledge is important. The troubleshooting side or the outlier side is more interesting. That’s what I mean by the edges. But if you have cross-domain knowledge, so you know how the billing system interacts with the fraud system, interacts with the I don’t know, data collection systems with something else, then that, that’s very valuable at the moment.

[00:15:54.20]
You also want to practise your escalation skills and be ruthless. When does a human have to be involved here? And if there’s a temptation to say, oh, well, the system’s recommending this, I’ll just do that, really put the brakes on and ask yourself the difficult, uncomfortable question: why? Why are we doing it? Why is this happening? Why? What are, what are the outcomes if we don’t do this? What are the outcomes if we do this and it’s wrong? I would get comfortable owning the edge cases, these outlier cases, and say, put your hand up for them. I know it feels like work, but I’ve got, I think it’s fulfilling. I think you’ll find it fulfilling, certainly more than just doing the same thing every, and answering the same question all the time. The handler needs to know how to escalate up into leadership, and then the leaders need to know how to delegate amongst a team of humans to resolve the issue. So there’s still the old management skills there. It’s just the management layer itself has gone.

[00:16:55.18]
Get better at prompting and not generic prompting. And this is a scaffolding prompt and this is an n-shot prompt, but how it specifically applies to your work. So know the constraints for your work. Know where the AI doesn’t know the constraints. So that you have to add them in. Get better at managing the agents. And there’s a whole range of things there. Whenever an AI agent goes off the rails, it’s invariably because it wasn’t given thinking controls and thinking constraints.

[00:17:27.25]
So knowing all the bits and pieces that go into that is, I think, pretty critical. And if you can’t audit the AI agent’s output, If you don’t, somebody else will. So get better at AI so you are the one who’s piloting the AI and auditing it and escalating it and that sort of thing. I want you to reframe your career. If the old ladder was a rep, a representative, and then a manager, and then executive or leader, the new, the new one is AI handler. Demonstrated expertise. Actually, not just demonstrated expertise, but demonstrated judgement. The ability to say, that’s a good decision, that’s a bad decision. I think governance. So accountability and risk and all those things that maybe you don’t normally think about too much. Now you need to. And the other one is you do think about them, but it’s instinctive for you because you don’t have to explain to an AI. But once you’re working with an AI, you realise it’s got no clue. It’s literally like it’s first day on the job and you know, why would you do that? That, that would put things at risk. Oh, okay. It’s, it’s really interesting. I would not at this point in time be optimising for the management rung. And so any courses you’re doing in that area, I think you might want to think about shifting it into a leadership job as opposed to management. Be careful. There’s all sorts of courses out there that call themselves leadership, but they’re management. It’s a bit of a furphy. Managers have to do some leadership and some management, full stop. But that’s splitting out now. Just be aware of that. Take the list you made in the last video with the AI-friendly side and the humans-only non-negotiable side.
Take the AI-friendly side and then make decisions about how you would audit the AI to make sure it was doing each of those tasks correctly. So. So if on the audit side you have that you have to summarise new regulations coming through, how would you audit those AI-generated summaries? What would you have on your checklist? It wouldn’t be enough to go, yeah, it looks okay to me. Make sure that you’ve got a way of going back into the senior leadership team and say, I’ve audited all the work the AI’s been doing. This is a repetitive issue I’m seeing. This is what’s going on. Because now you’re showing leadership and now you are showing that you belong on a team with AI because you are managing the AI, you’re piloting the AI. You’re not just outsourcing to AI and letting it try and do your work for you.

[00:20:15.20]
Anyway, I’ve been talking way too long. I hope that made sense. Specifically focus on the compression and expansion. Just keep thinking to yourself, as the organisation restructures, it will compress certain things like the management layer, Amazon firing 14,000 managers, and then it will expand in expertise area, OpenAI, Anthropic hiring good quality comms people. And then you’ve got other companies that are hiring entry-level graduates and you’ll see a lot of jobs, must know AI to be AI handlers.

[00:20:51.22]
So if you can start to think about the career ladder and proofing your career that way, then you can work with AI and not work for AI. Hope you found this useful. And don’t forget to stay human.

Note: she’s a long one, thank you for sticking with it.

Resources for AI Proof Your Career – Restructure and AI Career Ladder links

• International Monetary Fund (IMF) Gen-AI: Artificial Intelligence and the Future of Work https://www.imf.org/en/publications/staff-discussion-notes/issues/2024/01/14/gen-ai-artificial-intelligence-and-the-future-of-work-542379 (note the distinction between job displacement vs job transformation)
• McKinsey How AI is—and isn’t—changing the future of work https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-organization-blog/how-ai-is-and-isnt-changing-the-future-of-work (note: these are the very companies at risk of compression).
• ABC Australia Amazon cuts 14,000 jobs as it ramps up AI push https://www.abc.net.au/news/2025-10-29/amazon-culls-14-000-jobs-in-ai-push/105945212 (note: AI washing or real compression?)
• IBM restructure and hiring pause due to AI – Bloomberg https://www.bloomberg.com/news/articles/2023-05-01/ibm-to-pause-hiring-for-back-office-jobs-that-ai-could-kill
• OECD AI and Work (papers/topic) https://www.oecd.org/en/topics/ai-and-work.html
• Gartner on delayering middle management. Somewhere in here https://www.gartner.com/en/newsroom/press-releases/2024-11-06-gartner-it-symposium-xpo-2024-barcelona-day-3-highlights (survey: “In a quest to build dynamic leadership capacity, 56% of CEOs said they will use AI to de-layer most middle management roles within the next 5 years.”)

The shift is subtle but powerful. You move from “give me ideas” to “help me think through this business properly”. That means defining what matters (for most startups: revenue and stability), anchoring advice in your actual channels and customers, and forcing the AI to explore options, not just optimise one path. It also means adding tension – short-term vs long-term, growth vs sustainability – because AI tends to collapse complexity unless you explicitly hold it open. Like, it loves to collapse as if it was never given context and constraints!

And importantly, you don’t let it agree with you all the time. A good startup doesn’t need a cheerleader – it needs pushback. When you instruct AI to challenge assumptions, flag risks, and point out weak logic, it becomes far more like a co-founder than a content generator. You still make the decisions, but you’re making them with more visibility, more structure, and fewer blind spots. Plus, someone to argue with is always a good idea!

AI for Startups in Australia – all startups, not just AI startups – need custom instructions that mentor them through their business.

Transcript for AI for Startups – Custom Instructions

Hello, my name is Laurel Papworth, and in this video series we’re doing AI is my startup mentor. So how to set up an AI to partner with you in your business with all the usual caveats. This video is about custom instructions, which is how you set up the whole AI account, whether it’s ChatGPT, Copilot, Claude, Gemma, Google, whatever, by filling in the custom instruction box in your profile, that typically is the base prompt and the system prompt. And in this video, I’m going to give you how to work with your elevator pitch and the AI to get the questions. Answer the questions, write out their custom instructions using specific headings that will really make sure that the AI is fully supporting you and your business and isn’t just giving you generic answers, because you can get that from any, any book or blog post or whatever.We need specific things for this particular business. Okay, so, on with the show. So I would recommend that you take your elevator pitch and give it to the generative AI, put it into ChatGPT, and ask for it to ask you questions so that it can help you write custom instructions for ChatGPT. And you want to say custom instructions for Claude or custom instructions for ChatGPT. They’re all different, they have different lengths, and it can get confused. [00:01:40.23]

So really anchor that into the platform that you’re using. The questions that are going to come back after it’s read through your elevator pitch is the stuff that’s missing. Who’s your target audience? Where do you want to be in the next three to six months? A whole range of questions, but it’ll ask you those questions, and I’ve got some on the screen, but you don’t have to limit yourself to answering those. You don’t have to answer any. If you don’t want to, it’s fine. But effectively, customer instructions are the elevator pitch, but for AI. So I call it the thinking brief. You’re briefing the AI on how to be a thinking partner for you so that it can mentor you through your startup stuff. [00:02:25.08]

Once you’ve done that, I would suggest that you type in these headings and then ask the AI to create the custom instructions around these headings. The reason I do this is I’m defining the option space, like where I want to come in with my customer instructions, and then I’m defining the anchoring of those customer instructions because I really need the AI to be specific for my business and not some kind of a generic business mentor. It needs to be specific for my startup, so I give it a role in Scope. And I say you are a business partner helping me with my startup in healthcare, in sals, in education, whatever it is that you’re doing and give it enough detail so it can really help you with that. Also ask it to prioritise helping you on what to do next for growth and then specify growth as revenue if that’s important to you. Maybe you don’t care about money, you’re just doing this for fun. Good for everybody else out there. Ask it to help prioritise near term revenue growth. And I would suggest that you say I prioritise revenue growth and stability over quick wins. So you do a priority weighting in there. [00:03:50.02]

The next one’s called directional context. It just simply means the main theme and it’s your primary context for setting up the AI to help you. So you say I’m a niche business with limited time and resources and money. I guess that’s resources. I’m focused on getting clients, delivering revenue, delivering great service, whatever it is.
That’s like maybe your three main priorities. And you can say I want to prioritise short term traction over playing the long game. And I usually have something in there about wanting to be sustainable as well. And this is an issue with AIs. It doesn’t deal with that tension that every AI startup holds, which is between getting money in the door and then not being swayed into something that everybody else does that you have already tried and you don’t want to keep doing.
So you always. Humans hold tension really well. We sit with being uncomfortable pretty well. AI doesn’t. It’s always resolving and optimising.
So you gotta get it to hold its horses on some of those things. Add in needle facts. So add in needle facts. Needle facts are anchors. Tell it to assume or to ask for the target audience where most of the interest comes from.
Like you can tell it, most of my interest comes from LinkedIn. Or you can ask it to assume where that is coming from and then define it for you. Needle facts or anchors means make the advice specific to this point. So rather than saying come up with a marketing campaign, you want to anchor it into make come up with a marketing campaign for LinkedIn. In a normal prompt in a custom instructions, you’re just itemising very salient key points that will impact the answer.
I mean, maybe you like going for a jog, but unless you’re selling sports equipment, it’s not an anchor fact. It’s just kind of an interesting point about you. But if you do jog and you’re selling shoes, jogging shoes, Then put that in. It’s an anchoring point. The option space typically in a prompt is where you want it to come in at.

[00:06:07.29]
Like if you ask for, I don’t know, an AI strategy or come in at a tactical level or an operations level or strategic level. In the case of customer instructions, when looking up at your AI startup, if it’s like the landscape that you’re playing in and it’s specifically around what you provide, so you can say to it, give practical step by step, it’s called chain of thoughts. Give practical step by step advice across marketing, acquiring clients, regulatory issues, whatever the key things are that you think the AI can help you with. Now, some of these things you’re a subject matter expert on, so read through its work, just double cheque that it goes through to. It gets through your philtres and then the stuff you’re not a subject matter expert on, just double cheque, ask it, where possible, to use one output and then to regenerate for multiple uses.
So if it’s. If you’re working on a business plan and the business plan has a finance section and a marketing section and different things, once you’ve generated that business plan to your satisfaction, then you can, say, read through the marketing section and come up with 10 LinkedIn posts for me. Generate five images, whatever it is, you can use the primary document you worked on to generate again and again and again. That’s going to cut down your workflow substantially. So put it in the custom instructions that you want it to, like, recommend that it can do that, just to remind you it can, because I keep forgetting how bloody brilliant the AI can be when I’m monitoring it and watching it very closely under constraints.
This is the guardrails part you’ve probably heard people speaking about. Be careful because AI can drift. It will read the constraints and then go, I’m going to ignore them because there’s so much information or there’s so much stuff going on, I need to prioritise so many competing interests. But if you define the constraints in the custom instructions and then later on put the constraints in actual prompts, as you’re working through different projects, you’ll notice that it drifts away from those constraints. So one of them might be, you must comply with Australian regulations around advertising, and that way you won’t accidentally walk yourself into a brand reputation or into a regulatory hearing because the AI told you to do something that maybe you shouldn’t have done. [00:08:43.07]

I would also, in the constraints section, prioritise competitive pricing over something else so you can feel like you’re competitive from the beginning. Obviously, if you’re in luxury goods or you’re a senior strategist who’s just going out on their own and can charge a premium, you might invert that and say come in in the top 25% of my market when it comes to pricing, something like that. But put priority weighting in. You might say something like prioritise competitive edge over low pricing. It is possible AI will try and race you to the bottom with pricing.
Everybody else out there is $25. Why don’t you do $10? You don’t want that. You want to push unique selling points and competitive edge. And you, you’re brilliant.
Tell everybody that you’re brilliant and insist the AI does as well. I like to put format in custom instructions, even though I might change it later on and it will compete with how I’ve asked for the format to be versus the other. Let me give you an example. I asked for paragraphs. Don’t do I hate it when it’s one line, space line, one line, space line and it’s a toilet roll to read because it reminds me of those LinkedIn posts where they want to keep you engaged to play gamification with the LinkedIn algorithm which looks at dwell time.
I don’t. I’m not doing that with ChatGPT. Just give me prioritise three succinct, dense paragraphs over a long toilet roll. What’s the actual wording? Avoid long scrolling responses.
I think I actually initially wrote avoid long toilet roll scrolls and then it changed it to that one. Voice is also here. Make the voice Australian English. Don’t say write in Australian or you’ll get Kui Koba and G’ day Sheila. Ain’t nobody got time for that but friendly professional.
You a startup so you can ask for the voice to be savvy, personal or personable. You definitely don’t want vanilla and corporate. There’s enough out there in that space. But if I got here Tone should start a Start tone should suit a startup founder, savvy, curious and grounded rather than corporate or overly form. I think I said boring is batshit and it rewrote it for me.
So once I’ve got all those headings in and you can literally write the headings and then ask it to fill in those custom instructions yourself. We need to take the output when we’re ready, copy it and paste it into the custom instructions box for the overall account. Now in Copilot you can’t do this for the whole company you’d need to set up an agent and then share the agent with your marketing person, your HR person, whatever. But for a startup founder, where you’re the only full time person, maybe you’re outsourcing to a bookkeeper and outsourcing to a marketing person, bits and bobs, but primarily it’s for you. Make sure it’s in any of the AI tools that you’re using. [00:12:00.04]

Custom instructions at the end of the day are the base prompt and the system prompt. Who are you? What are you about? And then finally how you want the system to behave. So this is a bonus point for the next part of the custom instructions.
This is personal. So for some of you you need a positive reinforcement and others need to think things through a little bit more. So I have a thinking control called pushback. Do not assume my ideas are good or viable. We actually use the words that they have legs.
But you do you, where appropriate, challenge my assumptions, highlight risks and point out what may not work. Point out gaps in logic, misinformation, missing information.
After all, AIs got this massive brain and it’s thought of things that I haven’t even started thinking. Offer alternative approaches if something seems weak or unclear. And be constructive. I actually have to be gentle with me, be constructive, but do not default to agreeing to everything I say. I need a business partner.
I don’t need a yes man. It’s kind of annoying when AI wants to encourage me in things that I know just aren’t going to fly once I actually sit down, have a think about it. So it’s been very interesting the way it has pushed back on me from time to time, including buying some startup software I really wanted because I’m a geek and I want to play with it. And it said, yeah, no, you’re buying this for FutureLOL to implement in your business. And FutureLOL has a really poor attendance record.
I went ow. But it was true. So the pushback can be very helpful. And it, I think it saved me three or four hundred bucks, so that was worth it. The primary thing about pushback thinking controls is don’t just help me move forward, help me understand and mitigate possible mistakes.
You’re still the human in charge, you still make the decisions. But you do need somebody to say, are you sure about this? That’s what a buddy would do. So if AI is your startup mentor, the mentor is not going to tell you yes or no to do something, but will have a look on their face and go, are you joking right now? Like, I need that I need that for my AI as well. [00:14:46.08]

Obviously, if you’re the opposite and you’re negative about everything, or at least you’re extremely concerned and anxious and so risk averse that you might need to flip it and ask it to take a different approach. But hopefully that little bonus point, hopefully this pushback area really helps. So I hope you found this useful and I’ll see you in the next video. Thank you.

Resources for AI for Startups in Australia

• MIcrosoft has a hub for startups “Welcome to the AI for startups hub. Here you can find resources to help you integrate AI into your startup business. Browse through a collection of learning resources and FAQs to find the information you need.” https://learn.microsoft.com/en-us/startups/ then checkout https://learn.microsoft.com/en-us/startups/learning-path
• Of course that means ChatGPT/OpenAI have one too https://openai.com/business/learn/
• ScienceDirect: Generative AI for growth hacking: How startups use generative AI in their growth strategies https://www.sciencedirect.com/science/article/pii/S0148296325001432

If you have any good resources for Startups re: AI please let me know? This will be the greatest expansion area once AI automation hits white collar workers.

Transcript of Telos and AI

There’s been a few times that I’ve spoken about teleology and telos. It’s a Greek term, philosophical, meaning “purpose”. And with AI, it doesn’t have purpose. It has optimisation. It does not have the kind of purpose I’m talking about, which is capital P Purpose. So, dharma “this is what I was made for. This is what I’m meant to do”. And some of that comes with the fact that we are finite, so we look for meaning or purpose in a short time span. But a lot of it has to do with the fact that the intelligence of AI is artificial, so it does not have telos built in. Now, with humans, we have a bit of an argument about dharma and meaning and purpose. For some people… they don’t like philosophy. They sit inside their job or within their own business, and they’re just looking at revenue. And that seems fulfilling enough for them. I’m not talking about that. I’m talking about that where you’re willing to walk away from, everything, in order to fulfil what you feel like you have to do. And then telos will guide us. Individuals have telos, but there’s also, you know, “where’s humanity going? What are we trying to achieve?” And we battle all the time between the reality of who we are and then where we want to go. [00:01:32.10]

AI does not do that. AI reinforces the patterns that it sees. So if you tell it 10 times and you do a priority weighting and 10 times you say, I want you to prioritise following regulations over pricing, it doesn’t matter what it’s for, contracts, implementations, anything. If 100 times (prior) you have followed pricing, the pattern matching says to the “AI, yeah, pricing’s more important than regulations. Regulations is a nice to have.” No matter— no matter how hardcore you come down with your guardrails, it will drift into what it sees the pattern as being. Teleology typically with us is something that we don’t necessarily have or we haven’t achieved it. It’s tomorrow. Tomorrow I’m gonna go to the gym and be healthier. Tomorrow I’m going to eat better. Tomorrow I’m going to be, I don’t know, nicer to people. But it’s a goal and it’s, you’re leaning into progress and into something in the future. [00:02:43.11]

And I would also suggest that because AI doesn’t have temporality it doesn’t actually follow the arrow of time, which is what humans do, that it won’t understand being better tomorrow. It simply optimises today on the patterns it recognises today. So we end up in a situation where we are continually trying to tell the AI, don’t do as I do, don’t do as the patterns in the past have shown you. Try and be better than that. There are a few studies that have come out, I’ll list them on my website, where, um, researchers have tried to programme the model to have telos, to have this higher purpose. But, uh, I don’t know if it’s how that would work. You’d need an adversarial agent to continually ask it you know, why are you sitting there instead of going to the gym? Or why is it we have an inner voice that does that. AI doesn’t have that. It doesn’t have an adversarial agent unless we build one in. So yeah, telos is one of the things that human beings have, AI doesn’t have. And it would be extremely hard to construct it inside AI to tell it that it had a finite life and that it had to come up with some self-determination, ideas of how it wanted to be its better self. I hope you found that useful, and I’ll see you in the next video. Thanks.

RESOURCES FOR AI and TELOS

• Brian Cantwell Smith The Promise of Artificial Intelligence (Reckoning and Judgment) https://mitpress.mit.edu/9780262043045/the-promise-of-artificial-intelligence/
• Based on the work of Michale Bratman on AI agents. https://www.researchgate.net/publication/2596320_The_Belief-Desire-Intention_Model_of_Agency
• Victoria Krakovna (DeepMind) on AI being all about the reward and not about the game LOL https://deepmind.google/blog/specification-gaming-the-flip-side-of-ai-ingenuity/
• Stuart Russell Human Compatible (AI should remain uncertain, humans hold the tension) https://people.eecs.berkeley.edu/~russell/hc.html
• Shinn et al REFLEXION looped optimisation not awareness https://arxiv.org/abs/2303.11366
• Madaan et al Self Refine https://arxiv.org/abs/2303.17651
• Maybe Yao on Tree of Thoughts? Or Graph of Thoughts? https://arxiv.org/abs/2305.16582
• Constitutional AI – Anthropic https://www-cdn.anthropic.com/7512771452629584566b6303311496c262da1006/Anthropic_ConstitutionalAI_v2.pdf
• Andrew Ng (and Stuart Russell again) re: humans inconsistent with telos. https://ai.stanford.edu/~ang/papers/icml00-irl.pdf
• Geoffrey Irving, Paul Christiano, Dario Amodei (back when they were at OpenAI) on adversarial agents to help with telos https://openai.com/index/debate/

Exercise: How do we do priority weighting with competing priorities/outcomes/telos?

• Aristotelian telos (purpose, future orientation)
• Modern AI alignment (objective functions)
• Cognitive scaffolding (prompting, structured thinking)
• Governance systems (oversight, escalation)

Using examples like IKEA’s redeployment of customer service staff into higher-value interior design roles, this shows how organisations are already shifting from task execution to outcome ownership. Customer service procedures were compressed, while a new interior design bureau expanded – generating over $1 billion in revenue. As an employee, you are encouraged to map your own work into two categories: functions that can be delegated to AI, and functions where human judgement must remain.

Typical CSR split between AI functions vs Human functions. For the AI Handler role.

The core shift is from “doing tasks” to “owning outcomes.” This is where agency and leadership begin. Rather than waiting to be restructured, you can proactively redesign your role – working alongside AI as a manager of systems while retaining responsibility for decisions that matter. If you’re not auditing AI, someone else is. If you’re not handling AI, someone else is. The opportunity is to stay inside that loop – and to AI-proof your career by thinking strategically.

AI Proof Your Career – Employee tips – Australia HR and employment

Transcript of AI Proof Your Career – Job functions vs Job Description

Hello, my name’s Laurel Papworth and I mentor companies and government organisations through AI and emerging technology change. This video is about how to AI-proof your career as an employee. So it’s an employee-focused one. I wanted to make it personal. What does it mean for you and your job and what can you do? As the world is changing around you. AI restructuring isn’t just about cuts. It’s also about, uh, compression and expansion. And this AI accordion is where AI comes in and compresses certain layers in the organisation, but expands others. So some areas compress, some areas expand. And completely new greenfields are opened as well. [00:00:56.27]

Take IKEA. They introduced a chatbot called Billy, I guess after their famous Billy bookcase, that replaced a lot of functionality within the customer service field. So anything that required simple lookups like, when will my delivery arrive? Or what’s my order number? Things like that. An AI chatbot can take over. Instead of getting rid of hundreds, in fact, I think it was over 1,000 staff, they took that staff and redeployed them to a new area called the Interior Design Advisory. So these customer service reps had deep domain knowledge about design and creating bedrooms and lounge rooms. And so they put them in with AI, with tools, AI tools to help them, to help design spaces for customers. So it was a new area, interior design, made over $1 billion. It’s very clever. It repositioned their staff, allowed AI to take over certain functionality and then put the staff into an outcome-focused area, compressed, Customer service expanded interior design and allowed human value to shift to where human value was needed. [00:02:20.00]

So here’s what I recommend for you as an employee. Take your job, think of it as a series of job functions, like a, a, a functional workflow, and split those functions into two columns. The first column is the one you think is going to be compressed. Things you do that are repetitive that an AI can do. Summarising the weekly report, putting together a to-do list, anything that you think an AI probably could do with you having oversight of what it’s doing. You are the manager of the AI. [00:02:57.04]

And then the second one is where you stop and think, where do I place in judgement? Like, When do I stop and think, oh, I need to think about this one? Because in all likelihood, that’s the one you don’t want the AI to optimise for. You need human judgement. So that can be context, creativity, synthesis. Did I say judgement? Accountability. If an AI can escalate something, you can set it up to do that. So this isn’t things you escalate to your manager. This is the stuff where you stop and you make a judgement call and then you go ahead. So the first column is the procedural functions. Anything that’s repetitive, anything that you think AI can compress, anything you’re happy to shift to AI goes in the first column. And then the second one is the additional skills that the human has, which is human value-based. That’s the bit that’s going to expand. [00:03:58.09]

Once you see your job as a series of workflow functions and some of them you’re going to manage in AI and some of them you’re going to step in as the, as the leader in the space, you now have agency. You’re going to act with agency, not just as a worker, but as a leader because you’re taking responsibility for the outcomes. So that shift from doing the task to owning the outcome, that’s the shift into leadership. It’s where leadership starts. And it’s also how you stay relevant to an organisation that’s changing around you. Anyway, I wish you well in your AI explorations. Don’t wait to be restructured. Start restructuring what you do. And remember, stay human. I’ll see you in the next video. Thank you.

Resources for Human Resources in Australia:

• OECD AI and Work https://www.oecd.org/en/topics/ai-and-work.html
• New Skills and AI are Reshaping the Future of Work https://www.imf.org/en/blogs/articles/2026/01/14/new-skills-and-ai-are-reshaping-the-future-of-work
• McKinsey Agents, robots, and us: Skill partnerships in the age of AI https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-skill-partnerships-in-the-age-of-ai
• HDSR MIT Can We Predict What Jobs HR Will Take? https://hdsr.mitpress.mit.edu/pub/ppjz2dg9/release/2
• ARXIV (Cornell Uni, preprint) AI and jobs. A review of theory, estimates, and evidence https://arxiv.org/abs/2509.15265 [Department of Computer Science and AI Centre, University College London International Labour Organisation (ILO) Oxford Internet Institute, University of Oxford; Copenhagen Center For Social Data Science, University of Copenhagen
• ARXIV (Cornell Uni preprint) Beyond Automation: Redesigning Jobs with LLMs to Enhance Productivity https://arxiv.org/abs/2512.05659 (UK Civil Service study)
• Price Waterhouse Coopers (PWC) The Fearless Future: 2025 Global AI Jobs Barometer https://www.pwc.com/gx/en/services/ai/ai-jobs-barometer.html
• Australian Government Potential impact of Artificial Intelligence (AI) on the Australian workforce https://www.aph.gov.au/About_Parliament/Parliamentary_departments/Parliamentary_Library/Research/Issues_and_Insights/48th_Parliament/potentialimpactofArtificialIntelligence

Part 2 of “AI Proof Your Career” on surviving corporate restructures due to AI will be out shortly (April 2026).

What is Copilot inside Microsoft Suite?

Transcript of Microsoft Copilot lecture

[00:00:00.06]
Hi there. My name’s Laurel Papworth and I teach workshops on Microsoft Copilot and artificial
intelligence at universities, corporates, and government. I thought that today we would have a look at
what Microsoft Copilot inside 365 Studio Bing means to business and government.
[00:00:29.28]
So the term Copilot is used by Microsoft in 22 different instances. They’ve used it since 2019. It’s a
generic and descriptive term. It just means an assistant that does a bit more, like has integrations. So
an AI assistant is the original ChatGPT. You ask it a question and it doesn’t ask a question back. It
just gives you an answer and you leave it at that. Also Siri, Google, autocomplete functions, spelling
AIs, those sort of things are assistants. Copilot is used by Salesforce and IBM and Oracle and Google
Duet. So Copilot are used by lots of different companies to really mean the integration of AI so that
it’s more useful. As an example, there are out of the 22 different uses of Microsoft Copilot, and there’s
Copilot Designer and for images, and there’s Copilot for There’s Dynamics 365 Copilot, there’s the
Power BI one. I can’t remember, there’s lots of them. The 3 that I like to focus on are Copilot Chat,
which can sit inside the Microsoft Office home screen. Home screen. Pretty sure it’s free at the
moment.
[00:01:52.27]
Always check when my videos come out. Things have changed in the last week. But Copilot Chat is
just ChatGPT integration or GPT integration with SearchGPT functionality. So it has the web_browsing
plugin plus a bit more. You see it on the left-hand side. You go in there, you just ask questions, you’re
good. And if you have both Microsoft 365 Copilot and also Copilot Chat, then you’ll see there’s a
button at the top which says web and work currently. I don’t think that’ll last much longer, but it’s
there. Now the second one, which is Microsoft 365 Copilot, and that’s the one that comes inside
Microsoft Excel. So in class, for instance, we do pivot tables, but you can ask it to analyse data and
create dashboards. We have Microsoft Word, so instead of having to take text from Word, put it into
ChatGPT, have it edit it for you, then copy it back again. Now it’s built within Word itself. You can
access it in the top right-hand corner or the button above your document. There’s lots of other uses,
for instance, Pages and Canvas, but I really wanna just focus on the fact that the, you can open up
Teams and have it, the Copilot read create transcripts, things like that. Go into your OneDrive, create
PowerPoint presentations, yada yada yada.
[00:03:27.19]
So you can see the difference. I think ChatGPT sits over there on its own. It’s an AI assistant. Now
that it’s moved inside the Microsoft suite, it’s a Copilot. The third one is Copilot Studio.
[00:03:41.03]
Basically custom GPTs, little baby apps. Um, and Copilot Studio, I think they have a free thousand-
seater at the moment called the Copilot Viral something, Copilot Studio Viral. Anyway, I can take all
the compliance and regulations for my industry, put them into a SharePoint folder or connect into the
backend systems, and then create a little app so people who need to know regulations and
compliance can ask in plain English and get the responses. Just a reminder, GPT, the P stands for
pre-trained. So it’s like Encyclopaedia Britannica, the regulations and compliant document built inside
are a year out of date unless you’re using a custom GPT or RAG or something like that.
[00:04:32.09]
They’re the three main ones. I think the future, which is going to be AI agents and agentic workflows
where it does the work for you, is coming. It’s not here yet, so we’ll have to stay with the term AI
Copilot. When did it come out? 2017, I think.
[00:04:51.05]
So probably 10 years, and then we’ll be purely into AI agents. We won’t have to ask how to do
something. We’ll just say, create me a website, sign me up for a new bank account, get me a, um, a
business number, something like that. In my courses, I go through Microsoft Copilot and pass on the
tips and structured prompting and things that I see, what’s improving, what needs to improve. When
are we gonna get custom instructions and base prompts and system prompts inside the main
Microsoft 365 Copilot without having to use Copilot Studio, that kind of thing.
[00:05:34.09]
But anyway, if you’re interested, come along to the courses. Otherwise, in the meantime, I’ll see you in
the next video and don’t And forget to stay human.

RESOURCES:

• Microsoft Copilot learning hub for IT Pros and Developers https://learn.microsoft.com/en-us/copilot/
• Microsoft 365 Copilot https://adoption.microsoft.com/en-us/copilot
• Copilot Studio (agents) https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio

Courses on this topic range from 1/2 day to 2 days. Please enquire.

Directional Signals, Needle Facts, and Option Space: A Framework for Thinking About Markets with AI

Most people approach AI and the stock market in the same way they approach financial news: “What should I buy?” The problem is that both analysts and AI tend to surface the same obvious trends; the headlines everyone is already talking about. When AI answers that question, it usually produces directional signals: the visible momentum of the market. In the AI sector right now, that might mean companies like NVIDIA or Cisco — the names dominating media narratives and investor attention. But markets are shaped by shared stories, and when everyone follows the same narrative, those signals quickly become crowded and self-reinforcing.

A more useful way to work with AI is to treat it as a structured thinking partner rather than a prediction engine. In this video, I demonstrate a four-step framework for exploring investments with AI: starting with directional signals (the obvious trends), moving to needle facts (the overlooked suppliers and dependencies beneath the headlines), then exploring option space through scenario analysis, and finally testing your assumptions. The goal is not to outsource judgement to the model, but to use AI to surface hidden structures, challenge your thinking, and help you understand an industry more deeply before making any decision.

Video of AI and the Stock Market – Direction Signals, Needle Facts and More

Transcript of Beyond the Headlines: Using AI to Explore Investment Structures

DIRECTIONAL SIGNALS FOR STOCK MARKET (00:00:00.20) My name is Laurel Papworth, and I thought I’d go through with you how AI can help and will help in the future with the stock market and investment. I’m going to stick in my stream, which is about AI, and you can apply it, the structured prompt and the scaffolding, to any sector that you’re interested in. And it’s not about investment. This is about how to work with AI as a thinking partner. So a lot of people would ask AI, “Oh, there’s a lot of talk about AI at the moment. What stock should I invest in?” And AI will do what every, I guess, financial paper will do, financial media or any analyst article, anything like that. It will go with the headlines, and it will use what’s called directional signals for trends. What are the trends in investing in AI? And it’s going to come back and say, I don’t know, “NVIDIA. Very popular”. And the thing with the market is that it’s a shared narrative. So as the narrative moves towards NVIDIA, it becomes self-fulfilling. And then once the tide turns and it moves away, same thing. So a directional signal is a trend. Tell me what’s obvious. Tell me what the media is talking about. Tell me what the analysts are talking about that thing. If you don’t frame it within an analysis… If you don’t frame it within analysts and media commentary, then it’s difficult to move on to the next part. Which we’re going to do now.

NEEDLE FACTS FOR STOCK MARKET (00:01:48.12) So the second question that I would ask once it had come up with NVIDIA and maybe Cisco for their quantum computing networking stuff is called needle facts. And if directional signals is, tell me what the bleeding obvious is, the needle facts would be, tell me the stuff I’m not thinking about. Tell me the stuff that isn’t clearly stated but can be inferred. Tell me the stuff that doesn’t really appear except as a footnote in an addendum, like what am I missing? What’s being missed here? And this is where you find the little gems. So in the terms of AI, you might ask for something that’s second or third level. What specific non-visible structural dependencies are there that I could invest in? For example, which glass, ceramics, materials, specialty materials, provide the fabrication material? So it’s not the rare earth minerals, it’s the next level on. It’s the little that goes into a chip, that goes onto a card, that goes into a machine, that goes into a data centre, that becomes AI later on. There’s a company in Japan that makes a glad wrap, a saran wrap, and all of NVIDIA chips use it. So they’re a supplier, a vendor to NVIDIA. So there’s a flow on effect. If NVIDIA is doing well, these guys will do well as well. And in fact, I think they’re fully sold out for the next two years. And they have a waitlist. So does the company that makes the cleaning fluid for the chips. So does another company, Toto, that makes bidets, and they also make something else to do with fabrications. So once you’ve got directional signals, which is the framing of the main shift towards NVIDIA or Cisco or something like that, and then you have needle facts, which are there’s this substrata of an ecosystem of smaller companies that are doing extremely well.

OPTIONS SPACE (00:04:04.22) The next question is, what is the options space for the next three to five years? So we build our story, our narrative, the human narrative on prior information. And so does AI. AI uses tokens and data sets and stuff like that to build its story out. So the options base, which is the third one, is going to be looking at a tree of thoughts style prompt as opposed to chain of thoughts. You’re used to chain of thoughts. That’s tell me step by step how to get from here to here, step by step by step. Tree of thoughts are, give me the options. It’s an “if this, then that” type of scenario where you’re running along the branches saying, what if NVIDIA fails and it’s a bubble? What if utilities become overly regulated? What if this happens? What if that happens? And the AI is brain is big enough to hold all that stuff if you articulate the scaffolding of the prompt correctly. So what are the realistic scenarios that could unfold in the next three to five years? What if FAB construction slows down? What if geopolitical changes around regulations and tariffs and things like that impact these Japanese and Taiwanese companies that are making the bits and bobs that go into American, European, and Australian data centres, and what if there is a continued AI CapEx search or there isn’t. So it knows to start to play with the weighting, and you’re effectively using options-based to create tension because AI doesn’t hold tension. Humans do. So we’re creating a liminal space for the AI to hold tension and then to think through, “Oh, what if this happens?”

TESTING HYPOTHESIS IN STOCK MARKET AI (00:05:55.14) And then the fourth one is testing. I wish the people that pitch stuff to me would put this through, their pitches through the testing mechanism. But what if I’m wrong? What if I’ve got a basic assumption wrong? What if I’m missing something critical that should be in here? What If things change dramatically, what would invalidate my hypothesis that the substrata glass and ceramic companies that support AI will have a boom? What if the AI cycle cools? I’m not saying we head into another winter or bust, but what if for one reason or another, everybody starts to back off, gets a bit of cold feet? It happens. It tends not to happen TOO much in emergent tech. A lot of people talk about it. By the time there is the backing off. It’s so truly ingrained(normalised) that the market has stabilised. But testing your hypothesis and telling the to not support you, but to give you critiques and criticism, be gentle with me, though, is really super useful.

OUTPUT FORMAT (00:07:10.00) I would tell the model to keep each of those four sections separate. The directional signal, which is the headline stuff, the needle facts, which are the little known gems, the option space, which is the “if this, then that” tree of thoughts scenario building, And the fourth one, which is testing. I would say definitely keep them separate. We really want to go from trend noise to structural substrate, to scenario building, to testing our hypotheses. So that we’re not just saying, “I got a bit of pocket money, what should I invest in?” And we’re moving towards a much more risk averse foundational strategy.

AI MENTOR for INVESTING IN STOCK (00:07:59.04) On one last note I just want to remind people that Warren Buffett said he wouldn’t invest in any industry that he didn’t understand fully. I think he said that and airlines, but that if he didn’t understand the industry, he wouldn’t invest. And I understand that. I probably even agree with him. But if you use AI as a thinking partner rather than as a stock prediction oracle, or the Delphic oracle of stock markets, it will help you think through and understand better what you are investing in and what the risk assessments are, what the salience is of the headlines, all of that kind of stuff. Anyway, I hope you found it useful, and I’ll see you in the next video. Thanks. Bye.

This is not investment advice but how to consider thinking about investing with AI as a mentor.

Resources for beyond the Headlines: Using AI to Explore Investment Structures

• Harvard Business Review (Michael E Porter) https://hbr.org/1996/11/what-is-strategy
• Carlota Perez https://carlotaperez.org/ and books on Technological Revolutions and Financial Capital
• McKinsey The Economic Potential of GenAI https://www.mckinsey.com/~/media/mckinsey/business%20functions/mckinsey%20digital/our%20insights/the%20economic%20potential%20of%20generative%20ai%20the%20next%20productivity%20frontier/the-economic-potential-of-generative-ai-the-next-productivity-frontier.pdf
• Probably Richard Rumelt — Good Strategy / Bad Strategy book if you can find it.
• Warren Buffet https://www.berkshirehathaway.com/letters/1989.html (his comment on Peter Pan investors “I believe” is funny and relevant)
• Congress, Semiconductors competition and policy https://www.congress.gov/crs-product/R46581
• ARXIV Decision Trees for Decision-Making under the Predict-then-Optimize Framework https://arxiv.org/abs/2003.00360
• ARXIV On the Complexity of Decision Making in Possibilistic Decision Trees https://arxiv.org/abs/1202.3718

Youtube video on OpenAI vs Anthropic for Department of Defence/War USA

My take on the AI In Military contract.

Hello, my name is Laurel Papworth, and I want to walk through the recent issue between Anthropic and the United States Department of Defence, and then OpenAI’s response, which ultimately secured the contract.

Anthropic refused to allow certain uses of its model.

Specifically, they drew a line around mass surveillance of American citizens and fully autonomous agentic warfare. In other words, you cannot simply hand decision-making over to the system and say, “Do whatever you want.” They framed this as a matter of democracy, privacy, and accountability.

It’s a position many people would find understandable. But while Anthropic articulated the problem clearly, they didn’t offer a structural solution that met the Department’s operational demands. The Department responded strongly, indicating that if Anthropic maintained those constraints, its models could be excluded not just from Defence but potentially from other government departments.

On the same day, OpenAI stepped in and signed an agreement to operate within classified environments. What’s interesting is that OpenAI’s “red lines” were not dramatically different. They also stated no mass domestic surveillance, no autonomous weapons targeting, and no high-stakes automated decision-making. The difference was architectural.

OpenAI proposed a layered safeguards approach.

First, the models would run on OpenAI’s infrastructure (essentially within Microsoft Azure OpenAI Cloud) rather than being installed privately or at the edge. That means the provider retains oversight of inputs, outputs, telemetry, and execution context. There are audit logs. There is real-time monitoring. There are dashboards. It is not an invisible black box sitting inside a private facility.

Second, there must be a clear personnel escalation hierarchy, what in corporate settings we call an escalation chart, and in military settings resembles chain of command. Accountability is explicitly human. The AI may optimise toward a goal, but it does not hold judgement, shame, blame, or legal responsibility. You cannot sue a model. Responsibility remains with named individuals inside institutional structures.

This is the part many organisations underestimate. Automation does not eliminate accountability; it relocates it.

In high-stakes environments (particularly Defence) decisions escalate. An operator reports to a supervisor, who reports upward. Legal and ethical oversight sits within that chain. AI outputs must be tied back into those same authority structures.

Third, OpenAI emphasised what is often called the safety stack: prompt filters, behavioural constraints, output sanitisation layers, anomaly monitoring. These are standard components in system cards and AI integration documentation, but here they were contractually reinforced.

Critically, the provider retained discretion. If the model is pushed into prohibited domains, OpenAI can shut the service down. There are audit requirements. There are compliance reviews. There are penalties for breach. These are not soft “please comply” clauses; they are enforceable conditions.

That has implications. Cloud control concentrates power with the provider. Contractual enforcement is not self-executing; it requires monitoring. Safety mechanisms can be bypassed if humans decide to override them. And humans themselves can become the weakest link (not necessarily through malice, but through rubber-stamping, fatigue, or poor system design).

What this case surfaces is a larger tension many organisations are approaching. There is a strong push toward full automation. But systems optimise toward defined goals. Humans hold competing tensions: speed versus compliance, cost versus reputation, short-term gain versus long-term stability. AI does not naturally hold those tensions in the way institutions must.

Delegating routine tasks (even weaponised processes) does not mean delegating responsibility.

Technical constraints, institutional controls, contractual enforcement, and operational transparency form a governance bundle. Without that bundle, optimisation can drift beyond intended boundaries.

Initially, I saw this as Anthropic good, OpenAI bad. On closer inspection, Anthropic framed the problem clearly. OpenAI translated that problem into governance architecture that institutions could operationalise. In environments like Defence, where stakes are existential, that architecture matters.

Let me repeat: you do not outsource judgement to a system. You remain vigilant:

• You surface risk earlier.
• You maintain escalation paths.
• You tie automated outputs back to legal accountability.

That’s not a military issue. That’s an AI governance issue for every serious organisation.

Resources for Military AI (Department of Defence)

• REUTERS: Anthropic cannot accede to Pentagon’s request in AI safeguards dispute, CEO says https://www.reuters.com/sustainability/society-equity/anthropic-rejects-pentagons-requests-ai-safeguards-dispute-ceo-says-2026-02-26
• ASIS (security): https://www.asisonline.org/security-management-magazine/latest-news/today-in-security/2026/february/Anthropic-Refusal
• ANTHROPIC statement: https://www.anthropic.com/news/statement-department-of-war
• OPENAI statement: https://openai.com/index/our-agreement-with-the-department-of-war/
• ORACLE relationship with DoW (AI, data and infrastructure should be here): https://www.oracle.com/au/defense-intelligence/
• SALESFORCE AI and Military Defence: https://www.salesforce.com/news/press-releases/2026/01/26/us-army-department-of-war-missionforce-announcement/
• REUTERS on layered protections and red lines defence https://www.reuters.com/business/media-telecom/openai-details-layered-protections-us-defense-department-pact-2026-02-28
• WJARR paper on anticipatory models: https://journalwjarr.com/sites/default/files/fulltext_pdf/WJARR-2025-3505.pdf
• TechPolicy.Press https://www.techpolicy.press/a-timeline-of-the-anthropic-pentagon-dispute/

This post was lightly edited by ChatGPT for comprehension and betterer English.

AI News Australia – Microsoft 365 Copilot has Memories added (same as ChatGPT Memories).

AI NEWS: Microsoft adds Memories to Copilot

[00:00:00.00]
Hello, my name is Laurel Papworth, and this is AI news. So a quick copilot update. In October 2025,
Microsoft added custom instructions to copilot, so you could tell it how to respond and not just what
to do. Now, in December 2025, Copilot has added memory. This means it can reference previous
prompts and outputs to give more consistent answers over time. This matters because pretrained AI
models, GPTs, don’t remember anything, so they need a memory function. You can turn memory off if
you want fresh eyes. It’s under the personalization menu in settings. But for ongoing work, this will
make copilot far more useful. It now is more like a long-term colleague than an intern that started this
morning.
[00:00:51.17]
Just make sure they don’t steal the last muffin out of the breakout room. See you in the next episode,
session, whatever. See you then. Thanks.

Resources:
Microsoft Tech Blog https://techcommunity.microsoft.com/blog/microsoft365copilotblog/what%E2%80%99s-new-in-microsoft-365-copilot–november–december-2025

You will learn about the various use cases of AI, from large language models like GPT to assist you with copywriting and brainstorming, to applications like Dall-e and Microsoft 365 copilot which create imagery based on your input. Have GPT work with you to create strategies, and policies, campaigns and job descriptions, ads and blog posts as well as evaluate annual reports and spreadsheets/financials.

Upcoming Dates: this course has been run several times a month since March 15th 2023 including at Australian Institute of Management from 2023 until 2025. Please email pa@laurelpapworth.com for 28 July 2025

Executive Summary of the AI Is My Copilot Course

Boost productivity and efficiency with Microsoft 365 Copilot and AI tools like ChatGPT. This course provides a hands-on approach to integrating AI into your daily workflow, automating tasks, and making smarter business decisions. (see learning outcomes below for more information).

• Use AI and Microsoft 365 Copilot to streamline work in Word, Excel, PowerPoint, Outlook, and Teams
• Create custom AI instructions to tailor Copilot and ChatGPT to your role and business needs and projects
• Master advanced prompting techniques better inputs for better outputs (AI-generated results)
• Using Reasoning Models to write reports, strategies and indepth papers with citations.
• Train AI on your documents to extract and utilise key business insights – working with finetuned AIs.
• Generate AI-powered images and videos with DALL-E and Sora and Microsoft Designer Copilot
• Leverage private AI models for secure and controlled enterprise use
• Developing an AI strategy for the organisation. 

Who Is Laurel Papworth?

In 2023, Laurel Papworth is pleased to announce she has licensed her AI Is My CoPilot course at Australian Institute of Management (AIM) as she presents ChatGPT and other artificial intelligence tools at conference keynotes to large businesses, for the City of Sydney and other local gov entities as well as to small business as well presenting the workshop to Corporate and Government and NonProfits. AI Is My Copilot is a practical course on how to “centaur” or “copilot”, in other words, job share with AI, as a hands-on approach.

Her courses, since 2008, include artificial intelligence in algorithms and big data including F.A.I.R (Facebook AI), object recognition, behavioural AI and semantic AI such as Lexicon, Deep Face/Deep Text and Deep Mind.

In 2005 Laurel used her background in virtual currencies to submit a patent on decentralised currency for Play to Earn fitness using IoT, 3 years before Bitcoin. She taught Augmented Reality community building to Facebook, Google and Microsoft in California in 2011. Named by Forbes Magazine as Top 50 Most Influential Social and Online Community Expert, Laurel mentors companies visualising, building and managing Web3 and Metaverse solutions Artificial Intelligence solutions, merging technology, philosophy and social psychology to present a holistic Metaverse strategy.

Laurel Papworth consults with Australian Parliament House, Sony, banks, Channel Ten and non profits, she has 22,000 students studying her online courses and was Head of Community (and then Social) for virtual worlds like Sony games and Iron Will games since 2001 and taught “How to Build Massive Engaged Communities” in games and on web2 social media platforms at the University of Sydney from 2005.

More on Laurel on her podcast Alchemy of Innovation, YouTube, LinkedIn and her website laurelpapworth.com.

…and she regularly plays World of Warcraft, Roblox and other online MMORPG games with her avatar, SilkCharm.

HigherEd BIO Artificial Intelligence, Algorithms and Big Data – Australia and overseas:

Adult Education Trainer (Professional Development) in Social Media and other courses in the University of Sydney CCE from 2005 until 2017 (12 years continuous) Associate Lecturer in Bachelor of Design in the University of Sydney Faculty of Architecture in 2016 Lecturer in Creative Industries in the University of Western Sydney from 2012 Lecturer Post Graduate Communications in the University of Technology Sydney circa 2011 – 2014 Lecturer in Masters of Convergent Media course in the University of Western Sydney from 2008 -2010 Tutor in Computing in the University of Adelaide from 1989 – 1992 Lecturer AFTRS Lecturer MetroScreen SchoolGuest lecturer Universities in Singapore, Portugal, Effat/Jeddah (Saudi Arabia), community colleges and TAFE. Laurel also has a Cert IV in Training and Assessment.

PRIVATE TRAINING ON CHATGPT and AI for your Organisation: please contact me on 0432694992 or email pa@laurelpapworth.com for information on an inhouse/conference workshop or facilitated course.

PUBLIC TRAINING on CHATGPT and AI is primarily run through Australian Institute of Management (see below)

ChatGPT: AI Is My CoPilot course

My course is now licensed to Australian Institute of Management (delivered by me, Laurel Papworth).
Register your interest today (on the AIM website) for course dates.

Key Learning Outcomes Of ChatGPT and AI course:

Upon successful completion of this course you will:

• Gain an Understanding of AI Basics
  Develop a foundational knowledge of artificial intelligence, exploring different types of AI such as ChatGPT and Microsoft 365 Copilot. Learn how these tools are transforming industries including business, healthcare, and more, reshaping workflows and unlocking new possibilities.
• Explore Essential AI Tools and Technologies
  Dive into key AI technologies, such as ChatGPT for conversational AI and Microsoft 365 Copilot for productivity enhancement. Understand the mechanisms behind these tools and private large language models (LLMs). The course content is designed to be accessible for nontechnical professionals, ensuring everyone can confidently engage with these tools.
• Learn to Create Custom AI Instructions
  Discover how to design and save customised AI instructions using tools like ChatGPT and Microsoft Copilot Studio. These tools enable you to build profiles of key organisational data and create tailored, dynamic AI interactions that align with your specific business or job role.
• Master Advanced Structured Prompting Techniques
  Unlock the power of advanced prompting methods to frame questions and tasks in a way that maximises AI effectiveness. With tools like ChatGPT and Microsoft 365 Copilot, learn how to craft precise prompts that yield better, more accurate outputs by improving the quality of your inputs.
• Train the AI on Your Documents
  Learn to upload and integrate your organisation’s documents into AI systems using tools like ChatGPT and Microsoft Copilot Studio. These tools enable the AI to analyse, interrogate, and extrapolate information specific to your unique business needs, resulting in more customised and impactful solutions.
• Create AI-Generated Images and Videos
  Enhance your content creation skills by using tools like ChatGPT, DALL-E, and Microsoft Designer Copilot to produce visually engaging images and videos. These tools can transform your approach to marketing, presentations, and storytelling, making your content more impactful and professional.
• Increase Efficiency with Microsoft 365 Copilot
  Gain hands-on experience with Microsoft 365 Copilot—a powerful productivity tool that works seamlessly within Word, Excel, PowerPoint, Outlook, and Teams. Learn how to automate tasks, streamline workflows, and save time across your daily activities.
• Understand Private AI Models
• Explore the benefits of private AI models for your organisation. Learn how to maintain data privacy and control while leveraging advanced AI capabilities tailored to your specific requirements.

Typical Curriculum for 1 Day Artificial Intelligence Workshop

Understanding the technology

• We start the day by exploring essential AI tools such as ChatGPT and DALL-E. You’ll log in and learn how to use prompts, modifiers, and other AI clarifiers to get the results you want. Whether you want to create art, write a report, or generate code for a website, you’ll understand how AI learns from your inputs to refine its outputs, getting closer to your desired outcome. While this course is for non-technical people you will gain a solid foundation in the techstack required to build out an AI that can be trained on your business documents.

Structured Prompting – Better Input = Better Output!

• There are many ways to prompt using NLP (natural language processing) an LLM (large language model like ChatGPT). By understanding when and how to use the various types of prompts you’ll progress from asking a simple question to structuring sophisticated threads to get the output you want!

Exploring AI Tools for Business including Microsoft 365 Copilot

• You will receive a list of over 40 AI tools applicable to various business needs. While we won’t have time to cover all of them in class, we will prioritize a few key tools for in-depth exploration. From search to images to creating Powerpoint there are many options with integrating AI into your business.
• Custom GPTs allow various departments to create an AI that uses ChatGPT underneath but has their own documents. HR can create one, as can Customer Service as can Marketing. Let’s explore! Evaluate similar solutions for Microsoft 365 Copilot integrations.

Private LLMs and Risk and Regulation

• We will discuss privacy, plagiarism, and other ethical issues related to AI. This session includes an open forum to add and discuss additional concerns, exploring what can and cannot be controlled. Learn tips for engaging with government and regulatory bodies, and understand how to shape the future of AI collaboratively.

The Future of AI

• If relevant we explore how AI will impact jobs in your sector, tasks in your office, and the broader future of your industry. Building on our group discussion, you’ll use AI tools to generate a video, email, report, or image, applying the strategies and tactics learned during the course. This hands-on session helps you envision and prepare for the AI-driven future.

Alternative: Private one-on-one Single day in the Blue Mountains
for 1 to 3 people in your team

NOTE: if you are booking a private consulting day with me in the Blue Mountains during April 2023, please use THIS link after discussion with me. +61432684992. Limited number of days available!

Central Finance Office needs a proposal and/or quote? If you have any questions, want to run this privately for a conference or offsite, need a specialist Proposal etc, please email us with question, dates and detailed request. A GST Invoice/Receipt is issued if booked online, but sometimes Head Office needs more. Alternative call +61432684992

Not ready to join a course – watch my free videos!

• 

  Artificial Intelligence

  AI and Org Restructure – AI proof your career ladder

  ByLaurel Papworth

  Artificial intelligence is restructuring organisations by compressing routine work, reducing management layers, and expanding roles that require human judgement and accountability. Understanding where your work fits in this shift is key to staying relevant in an AI-driven workplace.

• 

  Artificial Intelligence

  AI for Startups: Leverage Your Custom Instructions #Australia

  ByLaurel Papworth

  Most founders don’t need more AI tools. They need better thinking. Or at least, someone to think with (if they’ve been brought in with information). This approach treats AI not as an answer machine, but as a business partner – one that is briefed properly, challenges when needed, and guides toward real outcomes like revenue,…

• 

  Artificial Intelligence

  Does AI Have Telos?

  ByLaurel Papworth

  Telos is our North Star – getting AI to develop a north star is… interesting. Can it be done?

• 

  Artificial Intelligence

  AI Proof Your Career 1 – Job functions vs Job Description – Australia HR

  ByLaurel Papworth

  I wanted this piece to reframe “AI-proofing your career” away from fear of job loss and toward understanding how work is actually structured. Jobs are not single roles – they are collections of functions, often called “task bundles”. AI doesn’t replace whole jobs neatly; it compresses repetitive, procedural functions while expanding areas that require human…

• 

  Artificial Intelligence

  What is Microsoft Copilot?

  ByLaurel Papworth

  What is MS copilot chat? How about 365 Copilot? What are Copilot Studio agents? Tutorial.

• 

  Artificial Intelligence

  Stock Market, Pick and Shovel, Artificial Intelligence #Australia

  ByLaurel Papworth

  A practical framework for using AI as a thinking partner in investment analysis: moving from market trends to hidden dependencies, scenario planning, and testing assumptions.

About Laurel Papworth

Artificial Intelligence

Laurel has taught AI, algorithms and big data at University of Sydney, UWS, UTS, UoA, AFTRs, and Universities in Singapore, Portugal and Saudi Arabia and currently teaches AI at Australian Institute of Management.

Educator

CERT IV Training and Assessment certified trainer (Diplomas and Certificates etc) Adult Education. Taught social media marketing at Uni of Sydney for 12 years. More Info

Clients

Laurel has consulted, mentored, spoken at conferences, run workshops and project managed clients including: Senators and Ministers (over 20 of them), Australia Parliament House, Westpac bank, Prime Minister of Malaysia, Royal House of Saud, multiple universities globally, NBC Universal Studios, Channel Ten Big Brother community, Ministry of Defence Singapore and more. Client List

Personally

Laurel has 22,000 online students. Laurel Papworth personally connects to 11 million followers online and has taught around 100,000 people in the last 10 years. Named by Forbes Magazine in the Top 50 Marketing Influencers globally, named Head of Industry, Social Media (Marketing Magazine) and in the Power150 Media & Marketing bloggers (AdAge) Testimonials

Artificial Intelligence Course Outline: 

Looking for a 1 Day Foundation or 2 Day Intensive workshop on the new artificial intelligence tools made available to business and individuals? Want information tools like Midjourney, Dall-e, Stable Diffusion, ChatGPT, – when to use them and how artificial intelligence can save you time and resources? Want a comprehensive list of the prompts and modifiers to use those tools to their optimum – and get the results you want?? Plus there are a hundred new custom GPTs (apps or plugins) and AI Addons being released a week – which ones are best for your organisation or business? This course is also aimed at you if you want to detect if AI has been used in a business report or student essay – there are AI detectors available! We cover the Future of AI in various industries from content creation to marking exams to predicting behaviours. Also how will artificial intelligence underpin the Metaverse/Web3 – From ChatGPT, Dall-e from OpenAI to MidJourney and beyond. This is an intensive course.