If anyone's going to replace me with AI, I'd rather it was me. So early 2026 I set out to do it.

I gave myself a tough brief: double my own output. Not tinker or play — double. I've spent thirty years delivering the newest technology, ERPs, Business Intelligence Systems, and transformation programs for banks, mines, airlines, government and health. I'm having a solid go to actively use the technology everyone's racing to switch on to make me twice as useful. AI works, but the answer I've found has not been as much to do with the tools. It is in where I use (or don't use) AI, what questions I ask, and especially what information I give AI to work with.

The commodity problem

The uncomfortable part, for anyone treating AI as an edge, is that the big models are a commodity. Your competitor licences the same ChatGPT, the same Claude or Copilot, points them at the same internet, or even at their own entire file system the same way you can. They get the same competent outputs, confident (and sometimes wrong) answers, with the same value add in effectiveness that you do. If everyone can buy the identical thing, get the same answer, and make the same document, it can't set you apart. Michael Porter worked this out a long time ago and AI hasn't changed it. Advantage comes from doing different things from your rivals or the same things differently, never from a tool everyone can buy.1 I am realising that the rush to switch AI on across every process, to throw all your data at it, to ask it every question, and to pour years of hard-won knowledge into someone else's training pipeline is buying the same horse as your neighbour and calling it a race.

If everyone can buy the identical thing, get the same answer, and make the same document, it can't set you apart.

Why before how

So I started somewhere else and I regularly return there. Not with the technology but with the question Simon Sinek would ask first: why. My why is personal. I want to be the one who makes me twice as useful, not let someone else's tool make me redundant. By regularly asking why my whole thinking has turned over. The more I've worked on "how do I double my value to a client", the more I'm seeing it is not to "use AI on more things."

Four months of failing first

My first day setting up my own stack I gave it every document I had. That failed. The next month was spent in tagging and metadata: a small improvement but no shift to the needle. Months two and three I used AI to architect, discuss, and code but still no breakthroughs. I'm slow, so it wasn't until month four I started to realise value wouldn't be in finding information others could Google, and now prompt for, or by building more, faster.

The lever is what you feed it

The lever to start to see value, it turned out, wasn't interchangeable access to the smartest models, best pipeline engineering and memory, or the most data. I have all of these and they all matter a bit but it has been the quality of what I give these to work from that's mattering more. I'm no longer feeding in everything I've collected and valued over multiple academic and business journeys. I curate locally the things that really matter and work. Around 1000 pieces, constantly growing with experience, engineered right but curated carefully. Steve Jobs famously transformed a failing Apple by having fewer products and more focus. Gordon Ramsay's shows turn around failing restaurants by reducing the menu. I've built a sovereign stack focussed over my own corpus. My methods, templates, decades of wiki reflections and academic "so-whats". The models bring the smarts and my focussed corpus brings the grounding. It all stays in local or sovereign models, private, not trained on. I can reach it from anywhere on any of my devices. The lesson that mattered has been that a small, carefully and constantly curated body of real expertise brings more ability to things a client actually pays for. Precision, authority, judgement, confidentiality. The same may be true for you and your enterprise.

The models bring the smarts and my focussed corpus brings the grounding.

Judgement, wrapped around AI

I don't think the advantage is the technology. Advantage is from wrapping AI in good data and human judgement. What you feed the retrieval augmented storage. How you curate and judge what goes in. The questions you ask. How carefully you vet the outputs, create quality loops, and put humans at the right place within this. Deeply understanding where your data is going. My AI lets me be myself, on purpose, and using tools to amplify what I already know rather than average into a commodity. The LLMs will get better and commoditise. My stack can already swap these in and out at will. My knowledge base and judgement are the parts that will last in value and grow as I build the right learning loops.

An honest scorecard

I'm at the start of measuring the gain, not the end of claiming it, so I've been cautious to write. The way I work has already changed. I'm still building my PM multiplier and chasing double, but I've become discerning about what I ask, where I ask it, and how I push back on the answers to see the sources used. I'm trusting my stack on some client work. It let me rewrite a 110-page project management framework in a day — aligned to global standards, contextualised to the client, at a quality I was proud of. That is normally a week of work. It was done privately: on IRAP-assessed infrastructure, every document held in Australia, nothing via a public chatbot.

The harder questions

If you're weighing this up for yourself the most useful thing is to learn for a quarter and sit with the harder questions first. Who do you want to be? Where's your real difference: the thing a competitor with the same tools still couldn't copy? What do you know that no one else does, and are you protecting it or quietly handing it over? Hopefully you are not uploading value to the web models in the name of speed. Whilst early adoption does have value, the winners over the next decade are unlikely to just be the fastest adopters. On Porter's logic they can't be as anyone with a credit card can adopt a commodity quickly. By Steve Jobs' logic, or Ramsay's, it won't be those who do everything. It'll be those who focus. The winners will be the ones who were clearest about who they already were and used AI to become more of that. Just like the winners have always been when the world transforms.

The tools are the same for everyone. You're not. The strategy is in the curation.