The AI Readiness Checklist: Six Things to Check Before You Start an AI Project

Before you start an AI project, check these six things. Not because AI is complicated (it isn't, not really), but because the businesses that get burned by AI investment almost always skip the same groundwork.

AI adoption is no longer a hard sell. Nearly nine in ten organisations now use AI in at least one business function. The harder question is what happens next. Only around a third of those organisations have moved past pilots to scale AI properly across the enterprise. The gap between "we tried AI" and "AI is delivering value" is not a technology gap. It's a readiness gap.

We've seen this pattern repeatedly with clients who come to us after a first AI project stalled. The model wasn't the problem. The data, the systems, the ownership, or the plan for getting people to actually use the thing usually was. This checklist covers the six areas worth checking before you spend a penny on your next AI initiative.

1. Do you have a real use case, not just an ambition

"We want to use AI" is not a use case. It's a mood. A real use case names a specific decision, task, or workflow that AI will change, who is affected by that change, and what the improvement actually looks like in numbers.

Vague ambitions produce vague projects, and vague projects produce vague results that are hard to defend at budget review. The strongest AI use cases share three traits: a clearly defined problem, a measurable outcome, and a process that already exists today, even if it's manual or slow. AI works best when it's improving something real, not inventing something new from a blank page.

Before you approve budget, write the use case down in one paragraph. If you can't, that's the first sign you're not ready yet.

2. Is your data actually trustworthy

This is the one businesses most consistently underestimate. AI systems are only as reliable as the data feeding them, and most organisations have far less confidence in that data than they assume.

Gartner research puts the average cost of poor data quality at $12.9 million a year per organisation, once you account for wasted effort, bad decisions, and compliance risk. Feed an AI model inconsistent, duplicated, or outdated data and it won't just underperform, it will confidently produce wrong answers at scale, which is worse than no answer at all.

Trusted data means data that is accurate, current, consistently formatted, and traceable back to its source. If nobody in your organisation can currently tell you where a given dataset comes from or how recently it was checked, that's worth fixing before the AI project starts, not after.

3. Can the systems that matter actually talk to each other

AI needs to reach the data and systems it depends on, and in most businesses those systems were never designed with that in mind. Finance sits in one platform, operations in another, customer data in a third, often with no shared structure connecting them.

The volume of data businesses are generating makes this worse every year. Global data volumes are projected to rise from around 182 zettabytes in 2025 to nearly 394 zettabytes by 2028, and most of that growth sits in systems that were never built to be queried together. An AI tool that can't reach the right data, in the right format, at the right time, will always underdeliver regardless of how capable the underlying model is.

Before committing to an AI project, map out which systems the initiative will need to draw from and check whether they can actually be connected without a lengthy, expensive integration effort. This is where a proper data architecture review earns its keep, because it tells you exactly what stands between your current systems and a working AI solution.

4. Who actually owns this, and are they equipped to

Every AI project needs a named owner with the authority to make decisions, not a committee that meets monthly. Ownership means someone accountable for the outcome, the budget, the risk, and the relationship between IT, the business function affected, and any external partner involved.

This is a genuine gap in most organisations right now. Deloitte's research into enterprise AI adoption found that a majority of boards still have limited to no working knowledge of AI, which makes it difficult for leadership to provide meaningful oversight of AI investment or challenge decisions made lower down the organisation. If nobody senior enough understands what's being built, accountability tends to evaporate the moment something goes wrong.

Before you start, name the owner. If that person can't describe the project's purpose, budget, and success measure in under a minute, ownership isn't settled yet.

5. Have you actually addressed security and compliance

Security is not a box to tick at the end of an AI project. It needs to be part of the initial design, because AI systems introduce risks that traditional software often doesn't: models trained on sensitive data, automated decisions that need to be explainable, and new attack surfaces that didn't exist before.

Deloitte's enterprise AI research found that data privacy and security is the single largest concern organisations raise about AI, cited by nearly three-quarters of respondents, yet only around one in five organisations currently has a mature governance model in place for the AI systems they're deploying. That gap between concern and preparedness is exactly where breaches, regulatory action, and reputational damage tend to originate.

Before launch, confirm who has access to the data the AI system will use, how decisions made by the system can be explained if challenged, and which regulations (GDPR, sector-specific rules, or emerging AI-specific legislation) apply to what you're building. A proper AI governance framework, agreed before deployment, is considerably cheaper than remediation after the fact.

6. Will people actually use it

The best AI system in the world delivers zero value if the people meant to use it quietly work around it. Adoption is the checklist item most often treated as an afterthought, and it's the one that decides whether the previous five points were worth the effort.

Adoption depends on training, on trust, and on people genuinely believing the tool makes their job easier rather than more monitored. It also depends on communication: telling staff clearly what the AI system does, what it doesn't do, and why it's being introduced. Businesses that treat adoption as a communications and training exercise, not just a technical rollout, consistently see stronger results from the same underlying technology.

Before go-live, have a plan for training, feedback, and iteration. Ask the people who'll actually use the system whether it solves a problem they recognise. If the honest answer is no, that's worth knowing before launch, not three months in.

Bringing it together

None of these six checks are complicated on their own. Together, they're the difference between an AI project that quietly becomes part of how the business runs and one that gets quietly switched off six months later.

The organisations getting genuine value from AI right now aren't necessarily using more advanced models than everyone else. They're the ones who did the groundwork first: a clear use case, data they can trust, systems that connect, an owner who's accountable, security built in from the start, and a plan for getting people to actually use it.

If you're planning an AI project and want a clear-eyed view of where your organisation actually stands, our AI strategy team can help you run through this checklist properly, and build the foundation your AI investment needs to succeed.