How to tell if your data is ready for AI

DEFINITION

What does "AI-ready data" actually mean?

"AI-ready" isn't a single yes/no state — it's six separate dimensions, and a business can be strong in some and weak in others. Knowing which is which is more useful than a generic verdict. Data is only one of the areas a full AI readiness assessment examines.

Accessible

You can actually get to it — export it, query it, connect to it — without a special favor from whoever built the spreadsheet.

Structured enough

Not necessarily a perfect database, but consistent enough that the same kind of information sits in a predictable place.

Consistent

Different systems agree on the basic facts — the same customer, the same order status, the same product name.

Complete enough

The fields that matter for the task at hand are actually filled in, not left blank for a large share of records.

Current

It reflects the business as it is now, not as it was when someone last updated the spreadsheet.

Permitted to use

You have a legitimate basis to use this specific data for this specific purpose — worth checking before, not after.

THE CHECKLIST

Can you run a real self-check this week?

This isn't exhaustive, but answering these eight questions honestly gets you most of the way to knowing where you stand — no tooling required, just a conversation with whoever owns the systems.

Can you export the data you need yourself, without asking someone else to run a special report?
Could a colleague who didn't build the system explain what each important field actually means?
If you looked up the same customer in two different systems, would the basic facts — name, status, contact details — match?
For the specific task you have in mind, are the required fields filled in for most records — not just the recent ones?
Was this data touched or reviewed in the last few months, or has it been sitting untouched for years?
Do you know, roughly, why this data was originally collected — and whether using it for AI fits that purpose?
If a client asked what happens to their data, could someone at the company give a straight answer?
Is there one person who'd be the point of contact if a question about this data came up?

A handful of "no" answers isn't a stop sign — it's a map of exactly what to fix first.

COMMON TRAPS

What are the common traps?

The same four patterns show up in most SMEs, regardless of industry — worth checking for even before running the full checklist above.

Data trapped in PDFs and inboxes

The information exists, but it's locked inside documents and email threads instead of a system that can be queried.

Multiple systems, no shared truth

The CRM, the accounting system and a spreadsheet each hold their own version of the same customer, and nobody has reconciled them — the kind of gap our AI integration work closes.

Undocumented tribal knowledge

One person knows how the numbers really work, which fields to ignore and which exceptions to make — and none of it is written down.

No clear owner for data quality

Everyone uses the data, but nobody is responsible for keeping it accurate, so small errors accumulate quietly.

BEFORE YOU USE IT

What should you ask about GDPR before using your data for AI?

This is a genuine legal question, not a checkbox — and it deserves an answer from someone qualified to give one, not a blog post. What's worth doing before you start is asking the right questions, ideally with your data protection advisor.

What is the legal basis for using this specific data for this specific purpose?
Is using this data to train or run an AI system compatible with the purpose it was originally collected for?
Does data leaving your systems and going to a third-party AI provider require a data processing agreement?
Is there personal data in here that isn't actually needed for the task, and could be removed or anonymized first?
Who internally is responsible for answering these questions, and have they actually been asked?

None of this is legal advice — it's the list of questions worth bringing to whoever handles data protection for your business, before the data goes anywhere near an AI system.

FAQ

Frequently asked questions

Do we need perfect data before starting an AI project?

No. You need data that's good enough for the specific task you're starting with, not a perfect, fully cleaned dataset. Scoping to one task and checking readiness for that task specifically is far more realistic than trying to fix everything first.

What if most of our data lives in spreadsheets?

Spreadsheets aren't disqualifying on their own. What matters more is whether the structure is consistent, whether the same kind of information sits in a predictable place, and whether you can actually export it. A well-kept spreadsheet can be more AI-ready than a messy database.

Should we clean all our data before doing anything?

No — scope the clean-up to the task in front of you. Cleaning everything before starting anything is a project with no natural end point, and it usually means the AI project never actually starts.

Who should own data readiness at a small company?

Usually whoever already has visibility across the systems involved — not necessarily a dedicated data role, which most 20-person companies don't have. What matters is that one person is clearly responsible for answering questions about a dataset, rather than the responsibility being unclear.

If you want an outside, structured read on where your own data stands against this list — not a generic audit — that's exactly what the AI Retrofit Check looks at. And once the data is ready, here's where it would actually save you time.