ChatGPT Changed the Question
The organisations now scrambling to write AI strategies are, in most cases, writing technology procurement documents and calling them strategies.
The Gold Rush Mentality
Five months after ChatGPT made generative AI a boardroom conversation, the pattern is already familiar. Senior leaders who had never previously expressed interest in machine learning are now demanding AI strategies. Chief Information Officers who spent the last three years delivering cloud migration programmes are being asked to pivot — to produce, within weeks, a comprehensive plan for how the organisation will exploit this new technology.
The urgency is understandable. The demonstrations are genuinely impressive. The competitive anxiety is real. But the response — the rush to strategy — is reproducing exactly the mistake that defined the first wave of digital transformation a decade ago: starting with the technology and working backwards to the problem.
The Strategy That Is Not a Strategy
The organisations now scrambling to write AI strategies are, in most cases, writing technology procurement documents and calling them strategies. They identify tools. They catalogue potential use cases. They estimate costs and timelines for pilots. What they do not do — because the technology arrived before the thinking — is ask the prior question: what decisions do we make badly, what processes fail our customers, what knowledge do we lose, and could any of those problems be addressed by this class of technology?
This is not a semantic distinction. A technology-led strategy selects a solution and then hunts for problems worthy of it. A problem-led strategy identifies the organisation’s most consequential inefficiencies and asks what — if anything — could resolve them. The first produces innovation theatre. The second produces transformation.
In my experience, the technology-led approach dominates because it is easier to commission. A vendor demonstration is concrete; an honest audit of organisational dysfunction is uncomfortable. Asking a large language model to summarise meeting notes is a crowd-pleasing pilot. Asking why the organisation generates so many unproductive meetings in the first place is a political minefield.
The Digital Transformation Precedent
We have seen this before. The early digital transformation programmes of 2012–2015 followed the same arc: technology fascination, rapid pilot proliferation, executive impatience, and then a slow, expensive recognition that the technology was never the hard part. The hard part was changing how people worked, how decisions were made, and how the organisation learned.
The parallel is almost exact. Organisations that led their digital transformation with technology — buying platforms before redesigning processes — spent years and significant capital arriving at the conclusion that the platform was not the problem. The resistance was cultural. The gaps were in data. The failures were in change management.
Generative AI does not change the fundamental logic of transformation. It changes the speed at which organisations can make the same old mistakes.
Generative AI is more accessible than its predecessors, which makes the temptation to skip the foundational work even stronger. When the technology appears to work out of the box — when a prototype can be built in an afternoon — the case for doing the slower, harder strategic work becomes almost impossible to make.
What Starting with the Problem Looks Like
An AI strategy that begins with the problem rather than the technology asks different questions entirely:
- Where does the organisation lose value through slow, inconsistent, or poorly informed decisions?
- Which processes depend on institutional knowledge that is fragile — held by a small number of people, poorly documented, or at risk of loss?
- Where do customers experience friction that is driven not by policy but by the organisation’s inability to retrieve, synthesise, or act on information it already holds?
- What data does the organisation possess that it cannot currently exploit — and what would need to change before any technology could make it useful?
These questions are harder than which vendor should we choose? They require honesty about organisational weaknesses that most leadership teams prefer not to examine. But they are the only questions that lead to strategies worth executing.
The Practitioner’s Dilemma
For those of us who work in transformation, the current moment is both exciting and deeply familiar. The technology is genuinely new. The organisational response to it is not. The pattern I have observed across every major technology wave — from ERP to cloud to data analytics — is the same: the organisations that invest first in understanding their own problems, and only then in selecting their tools, are the ones that realise lasting value. The organisations that start with the tool almost never go back to do the foundational work they skipped.
The window for getting this right is narrow. Once an AI strategy has been written, funded, and announced, it acquires institutional momentum. Questioning its foundations becomes career-limiting. The time to ask whether the strategy starts in the right place is now — before the pilots have been commissioned, before the vendor contracts have been signed, before the organisation has committed to a direction it chose for the wrong reasons.
The question ChatGPT changed was not can machines generate text? It was are we ready to be honest about what our organisations actually need? Most, so far, are not.