The problem
Without a shared framework, "using AI" means something different to everyone. One person may mean an occasional chatbot question; another may mean a redesigned workflow. People cannot see their own next step, managers have no common language for performance reviews, and leaders cannot compare progress across teams.
What was built
A maturity model that people use to see where they are, improve the way they do their daily work, and prepare for performance conversations. It has four levels and four dimensions, and it is applied by vertical and role, so a sales team and a finance team are each assessed against what good AI use means for their own work. It is designed to be practical: a tool, not a report card.
Four levels
Experimentation
AI use is individual and informal. Wins stay personal and learnings do not spread.
Workflow Integration
Specific workflows are redesigned around AI, with shared prompts and early measurement.
Integrated Intelligence
AI is woven into most core work. Someone owns AI capability, and results connect to business outcomes.
AI Operating System
AI is not a tool the team uses; it is how the team operates. The destination for every function.
Four dimensions
A function's level comes from where it lands across all four, not one or two.
What AI sees
From manual prompts with no context to live access to systems and knowledge.
What AI does
From one-off assists to end-to-end workflows with humans reviewing outcomes.
Who extends AI
From no one to building AI capability being a normal part of everyone's job.
How the org has changed
The hardest dimension: real maturity changes how a team is organized and decides.
How people move up: feedback and coaching
The model works through several tasks at once, not a single score. People assess themselves, gather feedback from peers and managers, and agree on one or two concrete changes to daily work. Coaching helps them practice those changes, and the next assessment shows whether the level moved. This is the mechanism that turns an assessment into progress toward the company's AI goals.
How teams use it
Design choices
- Observable signals, not opinions. Every level lists concrete signs a team can check against.
- Honest baseline without judgment. Naming where you are is the first step to improving.
- One destination for everyone. A defined end state gives every function the same target.
- Connected to action. The model links to functional roadmaps, governance, change management, and learning, so assessment leads to a plan.
Skills shown
Strategic framework design, AI adoption strategy, measurement design, change management thinking, and turning a broad ambition into something every team can act on.
Put it to work
The AI Maturity Review toolkit turns the model into a quarterly routine: an intake template, an evidence rubric, and a prompt builder for review notes and a path to L4.
See the Quarterly Review Toolkit →Described in general terms. Company-specific assessments and targets are intentionally left out. The model is one part of a larger framework with the Learning Strategy and the AI Fluency to Automation method.
The problem
A framework only matters if teams actually use it. Self-assessments tend to happen once, get scored on gut feel, and disappear. Leaders end up with no history, no reasons behind the scores, and no place for the problems a review turns up.
What was built
A lightweight tracker that turns the maturity model into a repeatable routine a team lead can finish in 10 to 15 minutes.
- Log a review: pick a level for each of the four dimensions, with a plain-language description of what each level looks like.
- Current snapshot: the latest level per dimension, movement since the last review, and the gap to a target you set.
- Trend over time: one line per dimension, so you can see the needle move, or not, review by review.
- Action register: stalls, regressions, and gaps become named, owned, dated commitments.
- Review history: every past review with its evidence, for a full trail.
- Backup and restore: export everything as a file, since data stays in the browser.
Design choices
- Evidence is required. A level without an observation behind it is a guess, so the form will not save without evidence. It also explains why a level changed later.
- Small on purpose. The action register is capped in spirit at five open items: if it fills up, you are flagging faster than you are resolving. The tool says so.
- Every action has an owner and a date. Otherwise it is just a note.
- Private by default. No account, no server. Everything stays in the browser, so a team can try it without a rollout.
- Targets per dimension. Not every dimension needs to reach the same level at the same time.
Skills shown
Turning a framework into a practical tool, product and interaction design, data visualization, lightweight front-end prototyping, and designing for honesty in self-assessment.
The interactive tracker is not public for now. It may become a future option, for example within AI Studio. The example outputs are illustrative and use roles, not people.