Teams have now approved AI budgets for three consecutive years, and the results are in. PwC’s 29th Global CEO Survey found that 56% of CEOs have seen neither revenue gains nor cost benefits from AI, while only 12% report both. An MIT study made headlines with an even starker figure: 95% of generative AI pilots showing no measurable return.
Having spent over 20 years in technology, and the last several helping organisations adopt AI, I can tell you the difference between the 12% and the 56% is rarely the technology. It’s what was evaluated before the money moved. That evaluation has a name: an AI readiness assessment. Here’s what it should actually cover.
What is an AI readiness assessment?
An AI readiness assessment is a structured evaluation of whether an organisation can turn AI investment into production value. It examines five areas together: strategy, data, technology, people and governance.
The word “together” is important here. Most assessments I see stop at infrastructure – they audit the cloud environment, tick the API boxes and declare the organisation ready. Then the pilot ships, adoption stalls, the data turns out to be unreliable, and nobody owns the outcome. Infrastructure readiness told you almost nothing about any of that.
A quick word on assessment tools, since plenty of vendors offer scored questionnaires. They’re a reasonable starting point. But a readiness score without an owner, a budget line and a sequenced plan changes nothing. Treat the tool as the beginning of the conversation, never the conclusion.
An AI readiness assessment framework that goes beyond infrastructure
These are the five areas we evaluate with clients, the key question behind each, and the warning sign that tells us to pause.
| Area | The question to answer | Warning sign |
| Strategy and value | Which specific problems is AI worth solving here, and what is each worth? | “We need an AI strategy” with no named use cases |
| Data | Is the data the use case depends on accurate, accessible and governed? | Nobody can say who owns the core datasets |
| Technology | Can AI connect to the systems where work actually happens? | Every integration requires a heroic workaround |
| People | Do teams have the skills and the appetite to change how they work? | Enthusiasm in leadership, silence on the floor |
| Governance | Are policies, oversight and regulatory exposure defined before launch? | AI usage is already happening with no rules at all |
Strategy and value comes first because it filters everything else. Readiness is always relative to a use case. An organisation can be entirely ready for AI-assisted document processing and nowhere near ready for autonomous decision-making.
People is the area leaders underestimate most. Adoption is a behaviour change, and behaviour change fails quietly. Teams accept the tool in the meeting and route around it in practice.
Governance has moved from optional to urgent for European organisations, with EU AI Act obligations phasing in. Knowing which risk category your use cases fall into belongs in the assessment, well before procurement.
PwC’s data backs up the multi-area view: CEOs whose organisations built strong AI foundations, including responsible AI practices and technology environments designed for integration, are three times more likely to report meaningful financial returns.
Why data decides most outcomes
If I had to weight the five areas, data would carry the most. Gartner predicts that through 2026, organisations will abandon 60% of AI projects that lack AI-ready data. The same research found that 63% of organisations either don’t have the right data management practices for AI or aren’t sure whether they do.
That matches what we see in client work. When an AI initiative stalls, the post-mortem almost always leads upstream of the model: inconsistent definitions across departments, records duplicated across systems, quality checked once a quarter when the model needs it checked continuously. None of this is glamorous work, which is exactly why it gets deferred, and exactly why it’s the best predictor of who succeeds.
How readiness gets proven in practice
An assessment on paper is useful. A scoped pilot in the real world is better, because it tests every readiness area at once, under production conditions, at limited cost.
Our long-term client Praxis, part of Maxeda DIY Group, is a good illustration of what accumulated readiness makes possible. When national research showed that consumers felt overwhelmed by home sustainability programmes, Praxis wanted an AI feature that gives personalised energy-saving advice from a photo of a room. Because the foundations were already in place, product data, integration paths, a team used to shipping, we took the Sustainability Check from decision to launch in four weeks, ahead of the peak DIY season. Over 30,000 users benefited from it between November and February, and optimisation work cut the upload-and-question flow from 19 seconds to 2.5.
Four weeks is what readiness looks like from the outside. This is also the logic behind The Idea Forge, the first phase of our Future Catalyst™ approach: identify where AI creates genuine value, then prove it with a tightly scoped, real-world build before committing serious investment. Readiness stops being a theoretical score and becomes something you’ve demonstrated.
Summary
An AI readiness assessment should evaluate strategy, data, technology, people and governance as one connected picture, because a gap in any one of them can sink the whole investment. Data deserves the heaviest scrutiny, since it decides most outcomes. Scored tools are a starting point, never a plan. And the strongest form of readiness evidence is a scoped pilot delivered under real conditions.
If you’re preparing an AI investment case and want an honest view of where you stand, our AI transformation team runs exactly this kind of assessment, and we’ll tell you plainly if the answer is “fix the data first”. Let’s talk before the budget moves.
