Senior HR and Technology leaders from within the financial services sector gathered to compare notes on closing the gap between AI ambition and workforce reality.

All attendees were tasked with driving the AI strategy for their organisation. Some had joined their organisations six months ago to find no AI strategy at all. Others had inherited 670 AI agents – six of which actually worked. One had received an internal audit report on day one and been asked to respond to it before they had even learned where the coffee machine was.

What emerged over the course of the evening was a striking convergence: the organisations moving fastest on AI were not the ones with the biggest technology budgets or the most sophisticated tools. They were the ones who had done the harder, slower work of answering a prior question: ‘What is our business strategy?’, before touching the technology at all.

AI strategy is only as good as your business strategy

Most organisations have the AI problem backwards. They start with the technology, deploying tools, building agents, running pilots, and then try to work out what it was all for. The results are predictable: low adoption, wasted budget, and a growing suspicion that the whole thing was oversold.

The fix is not a better AI tool. It is a better question. Not “what can we do with AI?” but “what are we actually trying to achieve, and can AI help us get there?” That shift sounds minor. In practice, it changes everything.

Technology without a strategy is just spending

Rolling out an AI assistant to every employee is not an AI strategy. It is a procurement decision dressed up as one. The tell is in what happens next: usage spikes for a few weeks, then quietly falls off a cliff. Emails get polished. Meeting summaries appear. Nothing about the business fundamentally changes.

This is not a failure of the technology; it is a failure of intent. A strategy requires a destination: grow this market, serve this customer differently, cut this process time in half. Without that destination, AI has no useful job to do. It becomes a solution in search of a problem, which is the most expensive kind.

The organisations that are genuinely moving forward on AI have done something harder and less glamorous first: they have defined what they are trying to build. Only then do they ask which tools might help.

The real blocker is almost never AI

Ask most senior leaders what is slowing their AI transformation, and they will say talent, or budget, or board buy-in. Dig a little deeper, and the same answer surfaces almost every time: data.

You cannot build reliable AI on top of unreliable data. You cannot automate business processes that have never been documented. You cannot personalise at scale when you do not hold the customer data you need. These are not AI problems. They are foundational problems that AI ambition has finally forced into the open – often for the first time.

This is, oddly, one of the most valuable things about the current AI moment. The aspiration creates the audit. Organisations that have spent years avoiding the uncomfortable question of whether their data is actually fit for purpose are now confronting it because they have no choice. AI starts the conversation. But it is data that determines whether anything comes of it.

Governance is an accelerator

The word “governance” tends to clear a room. In technology circles, especially, it carries the connotation of committees, slowdown, and bureaucratic overreach. That reputation is partly deserved, but it is also a symptom of governance done badly.

There is a meaningful difference between governance as a gate and governance as a framework. The gate model asks teams to pass through multiple approval stages before anything moves. It produces frustration, workarounds, and shadow deployments. The framework model gives teams a clear lane – here is what you can build, here is how it should be built, here is who is accountable – and then gets out of the way.

The evidence is fairly clear on which model produces results. The organisations scaling AI fastest are not the ones with the fewest controls. They are the ones with the right controls, designed to enable speed rather than inhibit it.

Culture change is delivered through outcomes

The hardest part of any AI transformation is not technical. It is human. People are not primarily afraid of AI in the abstract as they are afraid of what it might mean for their specific job, their specific skills, their specific place in the organisation.

No communication campaign fixes this. What fixes it is outcomes – visible, concrete evidence that the technology is doing something useful, and that the people using it are better off as a result.

This has implications for how AI programmes should be sequenced. Starting in the most resistant part of the business is rarely wise. Starting where the numbers will move, where the wins can be made visible, and where genuine believers can be found.

Building for the long game

Building AI capability internally cannot be outsourced, and the organisations that try will find they have paid for general principles while their competitors have built something they cannot buy: institutional understanding of how AI intersects with their specific processes, data, risk profile, and people.

The practical answer is a shared responsibility between the people function and the technology function, with learning and development leading delivery, and the technology function providing the substance. Neither works without the other.

But ownership is only part of the question. The deeper issue is what kind of capability is actually being built. The qualities that matter most, judgment, trust, relationships, adaptability, and the ability to read what an organisation needs and act on it, are not what AI competes with. They are what AI depends on.

The organisations pulling ahead are not the ones waiting to see how the technology develops. They are the ones doing the patient, unglamorous work: fixing the data, building governance that enables rather than blocks, growing capability from within, and creating the conditions for transformation to actually land. That work does not make headlines. But it is what separates the organisations that will look back on this period as the moment they pulled ahead from those that will wonder where the five years went.