Ask five people in most growing businesses what last month’s revenue was and you’ll get three or four different answers. Finance has one figure. Sales has another. A dashboard someone built eighteen months ago has a third. Nobody’s lying. The numbers come from different systems that were never properly reconciled, and everyone has learnt to trust their own version and get on with the day.
That’s survivable when the numbers only end up in a slide deck. It stops being survivable the moment you point an AI tool at them.
What is AI readiness?
AI readiness is a measure of whether a business can get reliable value from AI, not just run a trial of it. It covers four things:
- Data: is it accurate, complete, current and accessible?
- People: do staff trust the numbers, and can they use them without asking someone else to run a report?
- Process: what happens when a figure looks wrong, and who fixes it?
- Governance: who owns each dataset, and who decides which definition of a metric is the official one?
Most businesses assess the first one and ignore the other three. That’s why so many AI projects look great in a demo and fall over in real life.
Why do AI tools give confident, wrong answers?
Because AI doesn’t resolve disagreements in your data. It picks a version and states it with total confidence.
If your CRM says one thing and your accounting system says another, an AI tool has no way of knowing which one your finance team would sign off on. It will summarise whichever source it can reach, in clean, authoritative language. A wrong number with a polished delivery is worse than a wrong number in a messy spreadsheet, because people stop questioning it. We covered the same problem from the marketing measurement side in fix the data before you bolt on AI.
How do you check whether your data is ready for AI?
You don’t need a six-month project. A focused check looks like this:
- Map where your data lives and who really owns it. Not who owns it on an org chart from two years ago. Who touches it, edits it and gets blamed when it’s wrong.
- Pick the five to ten numbers leadership uses to make decisions. Revenue by channel, customer acquisition cost, retention, margin, whatever applies to you. Test each one against every source that feeds it. Wherever they disagree is your starting list.
- Check data quality and access. Is it complete and timely? Can the people who need it get to it without a workaround?
- Score readiness across data, people, process and governance. A simple one to five score on each is enough to show where the weak link is.
- Talk to the people who use the data day to day. The person manually reconciling two spreadsheets every Friday knows exactly where the cracks are. The monthly leadership deck won’t tell you.
If you can’t trace a number in your reporting back to a source you’d happily defend in front of your CFO, you’ve found your first job.
What does data governance mean for a growing business?
In plain terms: someone is accountable for each important dataset, everyone agrees on what each key metric means, and there’s a clear way to fix a number that looks wrong. That’s it.
You don’t need a committee, a framework or a dedicated team to start. You need a short list of your critical metrics, one named owner for each, and a written definition everyone has signed off on. Most businesses I speak to have never written that down, which is why “revenue” can mean three different things in one meeting.
Do you need a new platform to use AI?
Often, no. Not yet, anyway.
Plenty of businesses buy a platform to fix a problem that isn’t about the platform. If your systems don’t agree on basic figures, a new data warehouse or AI layer will faithfully copy that disagreement into a more expensive place. The platform decision should come last, once you know what needs fixing, what’s worth building first and what to leave alone. Sometimes the right answer is to tidy up what you already have.
Why do businesses skip this step?
Two reasons, and neither is about budget.
First, it’s less exciting than the AI conversation. Sitting finance and marketing in a room to reconcile a spreadsheet isn’t a great headline for the board.
Second, it’s uncomfortable. Admitting your leadership team can’t agree on five basic numbers feels like an admission of failure. It isn’t. It’s the normal state of most growing businesses, and the ones who say so out loud get to fix it first.
Frequently asked questions
What is the difference between data readiness and AI readiness?
Data readiness is whether your data is accurate, complete and accessible. AI readiness is broader. It includes data readiness, plus whether your people trust and use the data, whether you have processes to fix problems, and whether someone owns governance.
How long does an AI readiness assessment take?
A focused review of the main systems a business relies on, such as CRM, accounting, ecommerce, marketing and operations, can be done in around two weeks. A large enterprise with many more systems will take longer.
Does a small or mid-sized business need a data warehouse before using AI?
Not necessarily. Many businesses can get value from AI by first reconciling a handful of core numbers across the systems they already have. A data warehouse becomes worth it once those foundations are agreed and the volume or complexity justifies it.
What data problems most commonly break AI projects?
Conflicting definitions of the same metric, inconsistent naming across systems, data that is out of date by the time it is reported, and no clear owner when a number looks wrong.
Who should be involved in an AI readiness check?
Finance, operations, marketing and sales leads at a minimum, plus the people who prepare and use the reports day to day. Leadership alone will miss most of the real problems.