I read three AI maturity models cover to cover this month: one from IDC, one from AWS and Rackspace, and one from MITRE. Two of them are sales brochures wearing a framework’s clothes. And all three point at the same summit, which I think is the wrong mountain.
Here’s my take.
The IDC deck, in its own words, tells the reader: “Know your customer’s maturity level. Tailor your pitch. Win their trust.” Read that again. The audience isn’t the company trying to mature. It’s the vendor’s sales rep. “Maturity” is a qualification stage in someone else’s pipeline. The deck’s top level is the “AI-Fueled Organization”: AI-first strategy, continuous AI-driven business-model innovation. Higher maturity means more AI everywhere, which conveniently means more to sell you.
The AWS/Rackspace e-book is sharper and more honest about why AI stalls. Its best idea is real: the hardest moment in any AI effort is the jump from a pilot that works in a demo to a system that runs reliably in the business. Worth internalizing. But every capability in the book is an Amazon service, every stage advances “on AWS,” and the two calls to action are an AWS workshop and an AWS bootcamp. Same summit, too: “AI-first,” where AI becomes the core of business strategy. Great north star — if you’re an AI startup.
Then there’s MITRE’s model. It’s the rigorous one: vendor-neutral (government origin), built on the same CMMI lineage that the software world has used for decades, with an actual scored assessment behind it. And buried in its own text is the sentence the other two structurally can’t say:
“The target maturity level for an organization is a function of its mission and business practices.”
That’s the whole argument. The serious framework already knows the top isn’t always the goal. The sales brochures can’t admit it, because their summit is the product.
The mountain is the wrong one
Almost every AI maturity model on the market measures the same thing: how much AI you’ve adopted. Stage 1, you’re dabbling. Stage 5, you’re “AI-native,” “AI-first,” “AI-fueled.” AI is your strategy.
Ending there is like ending a tool maturity model at “you now use this hammer for everything.” It elevates the instrument over the job. “AI-first” is a fine goal for a company whose product is AI. For everyone else, which is almost everyone, the goals were never about AI. They’re revenue, margin, productivity, quality, speed, customer satisfaction. AI is one tool in service of those.
A company that uses AI in five well-chosen places and hits its numbers is more mature than one that’s “AI-first” and can’t show a result. The maturity that matters isn’t how much AI you run. It’s whether the AI you run gets you the results you came for, and whether you can prove it.
Measure outcomes, not adoption
So move the summit. The top of a maturity model that’s actually working for you isn’t “AI is everywhere.” It’s: AI dependably produces the business results you set out to get, you can prove it, and you keep reinvesting in what works.
Two things change once you frame it that way.
First, you define the destination before you grade the climb. Most models skip this. Before you ask “how mature are we,” answer two questions for each goal that matters: What are we trying to grow toward? (a real goal, not “adopt AI”) and How is AI supposed to help, and how will we know it did? (the mechanism and the measure). If you can’t answer those, you’re at stage one no matter how many AI tools you’ve bought. That gap is activity with no defined result. It’s the most common place I find companies stuck.
Second, you measure two things at every stage: progress and results. Progress means “is the practice maturing.” Results means “is the business actually better off.” A team can look more mature every quarter: more tools, more usage, more dashboards. And still deliver nothing to the bottom line. If you’re only tracking adoption, that team looks like a success. Track results too, and the work that improves the business shows up fast.
And then the part the vendors won’t tell you: the right target isn’t the top. Some goals deserve a fully closed, continuously-measured loop. Many are well served two rungs down. Aiming everything at “AI-first” is how you spend a fortune to move numbers that didn’t need moving. Pick the target that fits each goal. (MITRE agrees. It’s right there in the text.)
The honest version
The tool is genuinely amazing. I build production systems with it every week. That’s exactly why the question is never “how much of it do you use.” The question is “are you getting you the results you came for.”
If you’ve been doing AI for a year and you can’t answer that in a sentence a CFO would accept, that’s not a failure. It’s just the gap between adoption and outcomes. And it’s measurable. That measurement is the work I do: listen to understand your needs, name the goals, locate where you actually are, define the metric and the way you’ll track it, and tell you where to push next and where to stop spending. Vendor-neutral, because I’m not selling you the AI.
If that’s the conversation your leadership team needs to have, that’s the [AI Outcomes Assessment]. Reply and I’ll send the one-pager.
I’m Brian Fromme. I help mid-market teams get real results from AI. Built and shipped, whether that’s code or content. More at atelier.purpleblossom.ai.