You have not seen value from AI yet. Here is why, and what to do about it.

There are two reasons most businesses have not seen meaningful value from AI, and they are equally responsible for the situation you are probably in right now.

The first is that a lot of vendors oversold it. They promised transformation and delivered a chatbot that answers FAQs slightly faster than a search bar, sold you a platform with an AI badge on it that turned out to be a marginally smarter version of what you already had, and ran workshops that left your team knowing more about AI in general while being no better equipped to do anything specific with it. None of that was your fault. It was a genuinely oversold market and a lot of organisations got burned by people who were better at pitching than delivering.

The second reason is internal, and this one is harder to hear. Your organisation responded to the hype by creating the conditions that almost guarantee failure: a steering committee, a governance framework, a data readiness assessment, a policy review, and a standing agenda item in a monthly meeting where nothing ever quite gets decided because nobody in the room has the authority to start something small and see what happens. Both of these things have happened in parallel, and together they have produced a situation where two years have passed, budget has been spent, and the AI transformation is still described as being in progress.

Here is what to do instead.

Stop waiting for perfect data

Your data is not going to be perfect, it was not perfect before AI arrived and it will not be perfect when you eventually start, and every business that has built something genuinely useful with AI has done it with imperfect data, messy systems, and incomplete information. The AI does not need perfect data to be useful, it needs enough data to do something worthwhile, and you almost certainly have that already sitting in your existing systems.

The "we need to sort out our data first" position is comfortable because it is technically defensible and practically infinite. There is always more data to clean, more systems to integrate, more governance to establish before you feel ready, and if you wait for all of that work to be done before you start, you will be waiting for a very long time while your competitors are not.

Stop waiting for the committee to approve something

Governance has its place, and security, privacy and legal risk are real considerations that matter, but a monthly steering committee is not the right mechanism for moving at the speed AI development currently requires. By the time a committee has reviewed a proposal, sought input from the relevant stakeholders, reconvened three weeks later, and approved a pilot, the tool you were proposing to pilot has often been superseded by something better and the moment has passed.

The businesses making real progress with AI right now have given ownership to the people closest to the problem, not a centralised AI team, not a steering committee, but the operations manager who knows exactly where the bottleneck is, or the team leader who has been doing the same manual process for three years and knows every edge case and exception. These are the people who can identify what needs fixing, test whether a solution actually works in practice, and tell you honestly whether it has made a difference. Give them a budget, a clear scope, and the authority to make a call, rather than a ticket in a queue to a committee that meets monthly and decides nothing.

Start with one workflow

Not a strategy, not a platform, not a transformation program. One workflow. One process that currently requires a human to move information from one place to another, apply a routine rule, and pass the result along to the next step. Pick the one that costs the most, or annoys people the most, or creates the most errors downstream when it goes wrong, and build something that fixes it and see if it works.

The businesses we have worked with that have gotten the best results did not start with a grand vision for what AI would do for their organisation. They started with a specific, bounded, genuinely annoying problem and fixed it, and the broader vision came later once they had seen what was possible with their own eyes and built the confidence to go further. If you are not sure which process to start with, mapping where your time actually goes is usually the most useful first step.

Set a budget for experimentation and protect it

Twenty-two years ago, my marketing manager came to me wanting to spend money on something called Google advertising, and I was sceptical because it sounded speculative and I could think of better uses for the budget. But we agreed to allocate a thousand dollars and see what happened, and that thousand dollars was the beginning of a journey that eventually landed us in the BRW Fast 100. Not because a thousand dollars buys you very much, but because it buys you enough to find out whether something works, and once we knew it worked we knew exactly how to scale it.

AI experimentation works the same way. You do not need a transformation budget, you need a learning budget: a fixed amount, ring-fenced from the quarterly review, that a business unit can spend on trying something without having to write a business case every time they want to run a test. Some of what you try will not work, and that is entirely the point, because the things that do work will show you where to invest properly, and the return on those investments will be calculable in real numbers rather than projected in a deck that nobody quite believes.

Give ownership to the business unit

If AI sits with IT, it will be slow and carefully governed and will probably never quite fit what the business actually needs on the ground. If it sits with a central AI team, it will be theoretically impressive and practically disconnected from the real problems people are dealing with every day. If it sits with a steering committee, it will be endlessly discussed and never quite get started in a way that produces anything measurable.

Give it to the business unit instead, to the people who feel the pain of the problem every single day, because they are the ones who know what good looks like, they are the ones who will actually use whatever gets built, and they are the ones who have the most to gain from fixing it. Your job as a leader is not to govern the AI program from a distance. It is to remove the obstacles that stop your people from solving the problems they can already see clearly.

The window is not closing, but the gap is opening

You have not missed the boat, and the businesses that started two years ago have an advantage that is real but not insurmountable. What is true is that every month you spend in committee is a month your competitors are not spending there, and the gap between organisations that are moving and organisations that are still preparing is getting wider rather than narrower.

Pick one workflow, give someone the authority to fix it, set a small budget and protect it from the next round of cost cutting, and see what happens. That is how it starts for every business that eventually looks back and wonders why they waited so long to begin.