Your AI will only ever be as good as the data beneath it. Most companies' data isn't ready.

Every AI strategy has a glamorous half and a boring half. The glamorous half is the models, the demos, the use cases that make the board lean in. The boring half is the data underneath, and it is the half that quietly decides whether any of the glamorous part actually works. Almost everyone spends on the first and assumes the second. That assumption is where most AI disappointment is really born.

The scale of the gap is documented. Gartner found that 63% of organisations either do not have, or are not sure they have, the data management practices that AI requires, drawn from a survey of 1,203 data leaders. Read that again: nearly two-thirds are building AI on a foundation they have not confirmed is sound. And the report is pointed about why, because organisations that fail to grasp how different AI-ready data is from ordinary data management are putting the success of their AI at risk.

Isn't having lots of data the same as being ready?

No, and the confusion is expensive. Volume is not readiness. AI-ready data is data that is accurate, current, well-labelled, properly governed and genuinely fit for the specific job you are asking the AI to do. Most enterprise data is none of those things end to end: it is scattered across systems, inconsistently defined, patchy in quality, and assembled for purposes that had nothing to do with feeding a model. You can have terabytes of it and still not be ready, because the AI does not need more data, it needs trustworthy data.

This is why the foundation matters so much more with AI than with a dashboard. Point a capable model at poor data and it does not break in an obvious way. It produces something fluent, confident and plausible that happens to be wrong, and it does so at scale and at speed. The failure is invisible until it is costly, which is exactly the trap behind why most organisations fail at AI adoption: the tool works, the ground beneath it was never prepared.

Why does this belong to leadership, not just to IT?

Because data readiness is a strategic choice about where to spend and what to trust, not a technical chore to delegate and forget. Someone has to decide that the data behind your highest-value use cases gets fixed before you scale, that quality and governance are funded properly, and that data has clear ownership rather than being everyone's job and therefore no one's. Those are leadership decisions. Left to sort itself out, data readiness never happens, because it is nobody's exciting project and everybody's silent assumption.

How do I actually get AI-ready?

By treating the foundation as the first investment, not the thing you get to later.

1. Audit readiness, not volume. Ask whether the specific data behind a use case is accurate, current and fit, not how much of it you have. Volume flatters; readiness delivers.
2. Fix quality and governance before you scale. Poor inputs produce confident, wrong outputs. Sort the inputs first, then let the AI loose on them.
3. Give data an owner. Data without a named owner decays. Make someone accountable for the readiness of the data your key AI depends on.
4. Prioritise, do not boil the ocean. Fix the data behind your most valuable use cases first, rather than trying to clean everything and finishing nothing.
5. Treat it as continuous. Readiness is not a one-off project; data drifts and decays, so maintaining it is part of running AI, not a phase before it.

What does this change for me as a leader?

It changes the order of your spending. The instinct is to buy the AI and worry about the data later; the evidence says the return on the AI is capped by the data long before the model is the limiting factor. The organisations that get disproportionate value are not the ones with the cleverest tools, they are the ones that did the dull work of making their data trustworthy first, so everything they build on top of it can be trusted too.

That is the quiet discipline running through the end of business as usual: when the tools become universal, the advantage moves to the foundations underneath them, and data is the deepest of those foundations. Prepare the ground properly and AI becomes dependable rather than a gamble. It is one of the first things a board should be asking about before signing off the next AI initiative: not "which model", but "is our data ready to carry it".

About the author

Thomas Green

British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.