How AI maturity models turn gut feel into a plan

Why frameworks help

"How ready are we for AI?" gets a different answer from every person in the room. The CFO points at the last vendor invoice. The head of ops points at a project that shipped. The person doing the actual data cleaning has a much bleaker view. None of them are wrong; they're just answering different questions.

A maturity model doesn't settle the argument by fiat. It settles it by giving everyone the same dimensions and the same scale, so the disagreement becomes specific instead of vague. Instead of "are we ready?" you get "we're a 2 on governance and a 3 on analytics" — a claim you can check, argue about, and revisit next year.

The IBM Big Data & AI Maturity Model

IBM's model scores organisations 1-5 across dimensions including:

• Business strategy. Is AI tied to named business outcomes, or is it a slide in a deck?
• Information (data). Is data available, clean, and governed, or does every project start with a three-week data-hunting exercise?
• Analytics. Do you have basic reporting, some predictive work, or embedded decision automation?
• Culture. Do teams make decisions with data, or make the decision first and find data to support it afterwards?
• Governance. Who decides what AI is allowed to touch, and what happens when something goes wrong?

Level 1 is ad-hoc: things work because one person knows where everything is, and it breaks the day they leave. Level 3 is where most functioning organisations sit — some standard processes, some data quality work done, decisions increasingly evidenced rather than guessed. Level 5, optimising, means the organisation is tuning its own AI processes based on measured outcomes, continuously, without a project team having to be assembled first. Very few organisations are a genuine 5 on anything, and fewer still need to be.

A worked example

Take a mid-sized retailer wanting to build a demand-forecasting model. On paper the idea is sound. The readiness score tells a different story: business strategy is a 3 (leadership has said "use AI for forecasting" but hasn't defined what a good forecast is worth in stock-out reduction), information is a 2 (sales data lives in three systems that don't reconcile), and governance is a 1 (nobody owns the decision on what the model is allowed to override automatically).

That combination doesn't mean "don't do the project." It means the project as scoped will fail on data quality before the model ever gets judged on accuracy. The fix isn't a fancier model — it's three months reconciling the sales systems and one conversation to agree what the forecast is allowed to change without a human sign-off. Only then does the AI work have a foundation to stand on.

Pitfalls

The most common mistake is chasing level 5 everywhere. It's expensive, mostly unnecessary, and a distraction from the dimensions that actually block your priorities. A logistics firm doesn't need level 5 culture maturity to ship a route-optimisation model; it needs level 3 information maturity, because the routes depend on address data that's currently a mess.

The second mistake is treating the score as a one-off audit rather than a live input. Maturity moves — sometimes backwards, if a key person leaves or a data source gets retired. A score from eighteen months ago is a historical document, not a plan.

The third is scoring generously to make the leadership team feel better. A model is only useful if the scores are honest. If governance is genuinely a 1, say so; softening it to a 2 doesn't change the underlying risk, it just delays when someone discovers it.

What to do with the score

1. Identify which dimensions are actually blocking your prioritised projects — not all of them, just the ones relevant to what you're trying to ship next.
2. Sequence foundation work in those dimensions before committing budget to the AI work itself.
3. Re-score at least annually, and any time a major system, team, or leadership change happens.
4. Keep the scoring honest — evidence-based, not aspirational.

Where Florence helps

Florence's AI Readiness Assessment step uses the IBM Big Data & AI Maturity Model as its backbone. You score each dimension with evidence rather than gut feel, and the assessment carries forward into the Insights and Strategy steps, where it's used to flag which of your prioritised projects are realistically achievable now versus which need foundation work first. You're not just getting a number — you're getting a plan for what has to be true before the project you actually want to run will work.