The AI Operations Maturity Model: Find Your Stage, Know Your Next Move

A VP of operations told me his AI rollout failed. He'd bought a platform that deploys autonomous agents across departments. Six months later, adoption sat near zero and the contract was up for renewal he didn't want. The tool worked fine. His company just wasn't within two stages of being able to use it. He'd bought Stage 4 while his people were still at Stage 1, and no amount of onboarding closes a gap that wide.

This is the most common way AI efforts die, and almost nobody sees it coming. You can't buy your way past a stage you haven't lived through. The reason is boring and mechanical: each stage is a different unit of change, with different risks, and the skills from one don't transfer clean to the next. Skip a stage and you inherit its risks without the muscle you'd have built by earning your way there.

So here's the map. Five stages. Find yourself honestly, then make the one move that stage allows.

Stage 1: Chatting with AI

You open a chat window, ask it things, and sometimes the answer saves you twenty minutes. That's it. This is the prerequisite, not a stage we teach, because if you've used ChatGPT to draft an email you're already here.

The tell: your AI use is a series of one-off conversations that leave no trace. Close the tab and nothing changed about how work gets done tomorrow.

Next move: stop treating each prompt as disposable. Pick one task you do every week and write down the exact instructions that made the AI do it well. The moment you save that, you're building toward Stage 2.

Stage 2: Automate Myself

Now you've turned your own repetitive work into something repeatable. You have prompts you reuse, maybe a few connected tools, and a task that used to take an hour takes ten minutes every time, not just the time it happened to work. The unit of change is you. One person, your own workflows, your own judgment checking the output.

This is where individual leaders and practitioners live, and it's further than most companies get. A recruiter I know built a screening process that reads every inbound resume, scores it against the role, and drafts her rejections and callbacks. She runs it daily. It's hers, it's reliable, and she can explain every step.

The risk here is small and personal. If the AI messes up, she catches it, because she's the only one depending on it.

Next move: document what you built well enough that a second person could run it. That documentation is the bridge to Stage 3, and writing it will show you how much of your process lived only in your head.

Stage 3: Automate My Team

You take what works for one person and make it work for a group. This is the stage where most efforts quietly break, because the unit of change stops being you and becomes other people. Your teammates have different habits, different tolerance for weird output, and different reasons to ignore a tool you love.

A support manager rolls out a shared assistant that drafts ticket responses for her whole team. Half the team adopts it. The other half keeps doing it the old way because the drafts don't match their voice, and now she's managing two processes instead of one. The technology was never the hard part. The hard part was that automation now depends on people who didn't build it and don't trust it yet.

New risks show up here too. One person's mistake stayed contained. A team's mistake, repeated across every ticket, becomes a pattern customers notice.

Next move: stop optimizing the tool and start building the adoption. Find the one teammate who resists it most, sit with them, and fix the specific thing that makes it useless to them. Their objection is your roadmap.

Stage 4: Automate My Company

Now automation crosses departments and touches systems of record. Sales talks to finance talks to fulfillment, and the AI moves work between them without a human relaying it. This is senior leader and C-suite territory, and the unit of change is the organization itself: its processes, its data, its risk posture.

This is exactly the stage my VP tried to buy. The failure wasn't the platform. It was that company-wide automation assumes company-wide readiness, and his people were still saving prompts in a doc, if that. Cross-department agents need clean data, defined ownership, and teams who already trust automation at their own level. He had none of it, because he'd skipped the three stages that build it.

The risks are now existential to the business. A bad automation between finance and fulfillment doesn't waste twenty minutes. It ships the wrong thing to the wrong customer and bills them for it.

Next move: if you genuinely have teams operating at Stage 3, pick two adjacent processes and connect them, with a human owning the seam. If you don't have Stage 3 teams, this move isn't available to you yet, and pretending otherwise is how you become a cautionary tale.

Stage 5: Automate Any Company

You've done this so many times that the method transfers. You can walk into a business you've never seen and take it up the stages, because you understand the mechanism, not just your own case. This is where certified practitioners and licensed partners operate, and it's a different job entirely: you're automating the act of automating.

Almost nobody reading this is here, and that's fine. It's a destination, not a starting point.

Locate Yourself Honestly

The trap is that everyone rounds up. Ask people their stage and they answer with their ambition, not their reality.

So don't ask what you're trying to do. Ask what's running right now, today, without you touching it. If the honest answer is "a few good chats," you're at Stage 1, and that's a fine place to start building from. The expensive mistake isn't being early. It's buying for a stage you only wish you were at.

Find the stage where the work actually lives. Then make the one move it allows. That's the whole method, and it's slower than the pitch decks promise, which is precisely why it works.