At some point in 2026 a leader above you - a CPO, a VP of Product, a CEO who read one McKinsey slide - will ask a version of the question you have been dreading: "Where are we on AI?" If your answer is a feeling ("we're behind," "we're doing okay," "the team is anxious"), you have already lost the conversation, because a feeling cannot be funded, defended, or compared to last quarter. This lesson gives you the instrument that turns the feeling into a defensible artifact: a maturity assessment that audits your design team across six dimensions, scores each on a four-level scale, renders the result as a heat-map an executive can read in ten seconds, and produces a maturity report with named gaps and a 90-day priority list. It is the first thing you build as a Design Strategist, because everything else in L4 - the thesis, the roadmap, the vendor decisions - is downstream of an honest answer to "where are we, actually."

Why a Feeling Loses the Room

The reason "we're behind on AI" fails as a leadership answer is not that it is wrong. It is that it is unactionable and uncomparable. A leader cannot allocate budget against "behind," cannot tell whether you got better or worse since last quarter, and cannot distinguish your honest self-assessment from the design lead two floors up who said "we're crushing it" and meant "two of my people use ChatGPT." When design speaks in vibes while finance and engineering speak in numbers, design loses the resourcing argument by default, every time, because the org optimizes what it can measure and ignores what it cannot.

The maturity assessment is your translation layer. It takes the thing you actually know - which is a great deal, because you live inside the team - and renders it in the form the org respects: a structured score across named dimensions, with evidence behind each score and a prioritized plan attached. It does not make you more certain than you are; a good assessment is explicit about its own confidence. What it does is make your certainty legible, so that "the design-system dimension is at Level 1 and that is the single highest-leverage gap" becomes a sentence a CFO can act on rather than a worry you carry alone.

There is a second, quieter reason to build it. The assessment protects you from the two failure modes that get design teams hurt in a layoff cycle: overclaiming and underclaiming. Overclaiming ("we're fully AI-augmented") invites the cut, because if you are already transformed, why do you need the headcount. Underclaiming ("we've barely started") invites the cut from the other direction, because a team that has not adapted looks replaceable. A calibrated assessment - here is where we are strong, here is where we are weak, here is the plan to close the gap - is the posture that survives, because it reads as a team that is managing the transition deliberately rather than being managed by it.

The Six Dimensions You Audit

A maturity assessment is only as good as its dimensions. Score too few and you flatten real differences; score too many and the heat-map becomes noise no executive will read. Six is the number that holds, and these are the six, chosen because each fails differently and each is owned by a different lever you control. Audit all six and you have a complete picture; skip one and you will be blindsided by it within a quarter.

People

This is the capability and confidence of the humans on the team: who can run an AI-augmented research synthesis and verify it, who can brief a builder tool without losing authorship, who can audit a generated mock for the behavioral errors, and - just as important - who is anxious, who is resistant, and who is quietly hoarding AI skills as job security. People is the dimension most leaders ignore and the one that most reliably determines whether everything else works, because tools do not adopt themselves. Score it on distribution, not average: a team where two people are expert and eight are frozen is not "intermediate," it is a two-speed team with a specific, dangerous gap.

Tools

The actual stack: what is licensed, what is shadow-IT (people expensing their own ChatGPT Plus), what is integrated into the workflow versus opened occasionally in a panic. The failure mode here is not "too few tools," it is incoherence - five overlapping image models, three prototyping tools nobody agreed on, and no decision rule for which to use when. A team can be tool-rich and maturity-poor. Score this on coherence and governance, not count.

Workflows

Whether AI is woven into how the work actually gets done or bolted on as an occasional shortcut. At low maturity, AI use is individual, undocumented, and invisible - one designer uses it heavily, nobody knows where or how, and there is no shared pattern. At high maturity, the team has named, documented, human-in-the-loop workflows for recurring work (research-to-synthesis, design-to-code handoff, accessibility audit) with the verification step explicit. Score this on whether a new hire could learn the team's AI workflow from documentation or only by osmosis.

Design Systems

The single most leverage-positive dimension in 2026, and the one most likely to be at Level 1 on a team that thinks it is advanced. This is whether your design system is AI-readable: tokens in a structured format, components documented so an agent can read their intended use, a Storybook or equivalent that an MCP-enabled tool can consume without hallucinating props. A team with a beautiful Figma library and no machine-readable token layer is at low maturity here regardless of how good the library looks to humans, because the AI tools cannot use it, which means every generated artifact drifts off-system.

Brand Systems

The visual-and-voice equivalent: whether brand distinctiveness is encoded somewhere AI can respect it (brand-anchor reference sets, a documented voice-and-tone, tokenized brand variants) or whether it lives only in the heads of two senior designers and collapses into generic slop the moment they are not in the room. Brand systems is the dimension where "the average" is the enemy, so a team that delegates brand work to AI without an anchor is at low maturity even if the output ships fast.

Design Ops

The connective tissue: provenance logging, IP and indemnification policy, disclosure practices, the verification-tax accounting, and the governance that keeps the other five dimensions from each becoming a liability. Design ops is the dimension that turns individual capability into organizational capability. A team where every designer verifies their own AI output but nobody logs provenance is one legal review away from a problem, which is a design-ops gap, not a people gap.

The Four-Level Scale That Makes Scores Comparable

Each dimension scores on the same four-level scale, because a shared scale is what lets you put six different things on one heat-map and lets a leader compare this quarter to last. Borrow the shape from capability-maturity models the org already trusts, so the format reads as rigor rather than invention.

Level 1, Ad Hoc. AI use is individual, undocumented, and unmanaged. It happens, but by accident and in pockets. There is no shared pattern, no policy, no measurement. Risk is uncontrolled because nobody can see what is happening.

Level 2, Emerging. Some shared practices exist - a couple of documented workflows, a partial tool decision, a first stab at a provenance log - but they are inconsistent, optional, and dependent on specific individuals. Remove the one person who drives it and the dimension reverts to Level 1.

Level 3, Defined. The dimension has documented, team-wide practices that survive turnover: a tool decision rule everyone follows, a workflow a new hire can learn from docs, a design system with a real machine-readable layer, a governance policy that is actually enforced. The practice is institutional, not personal.

Level 4, Optimized. The dimension is measured, reviewed, and improved on a cycle. The team does not just have a workflow; it tracks the verification tax on it and tunes it. It does not just have an AI-readable design system; it monitors drift and remediates. This is the level most teams should aspire to on two or three dimensions, not all six - chasing Level 4 everywhere is a way to spend a year and achieve nothing shippable.

A word on honesty in scoring, because the temptation to round up is enormous and fatal. Score the dimension at the level the median person and the documented practice support, not the level your best designer demonstrates on their best day. If the workflow lives in one person's head, it is Level 2 at most, no matter how good that person is, because the org capability is what survives them leaving. The assessment is worthless the moment it becomes a marketing document; its entire value is that it tells you the truth you can then act on.

Score the median and the documented, not the best designer on their best day. A practice that lives in one head is Level 2 no matter how good that head is, because what survives turnover is the only capability the org actually has.

Running the Audit Without It Taking a Month

The assessment has to be cheap enough to run quarterly, or it becomes a one-time consulting exercise that goes stale. Budget about a week of part-time effort, structured in three passes. First, evidence gathering: pull what already exists - the tool inventory from finance's SaaS spend, the workflow docs (or their absence), the provenance log (or its absence), the design-system repo. Half your scores are answerable from artifacts that already exist or conspicuously do not.

Second, a short structured interview or survey across the team, six questions mapped to the six dimensions, anonymous enough that people tell you the truth about their anxiety and their actual usage rather than the answer they think you want. The people dimension in particular cannot be scored from artifacts; you have to ask, and you have to ask in a way that surfaces the frozen and the resistant, not just the enthusiastic. The distribution of answers is the score - if responses cluster bimodally (a few experts, a long tail of non-users), that is a two-speed team and you score people no higher than Level 2 regardless of how high the experts are.

Third, your own calibration pass, where you reconcile the artifacts and the interviews into a single score per dimension with a one-line evidence note attached to each. The evidence note is not optional; it is what makes the score defensible when someone challenges it and what makes next quarter's comparison meaningful. "Workflows: Level 2 - research-synthesis workflow is documented and used by four of ten ICs; no other workflow is documented" is a score you can defend and re-measure. "Workflows: 2" is not.

The Heat-Map: Making Six Scores Readable in Ten Seconds

The heat-map is the executive-facing surface of the artifact, and it does one job: let a leader see the shape of your maturity at a glance and let you see where to point your attention. Render the six dimensions as rows and the four levels as a colored scale - red at Level 1, amber at Level 2, light green at Level 3, deep green at Level 4 - so the pattern of color is the message before anyone reads a word. A team that is green on tools and people but red on design systems and brand systems has a specific, legible story: capable people with good tools who will produce drift at scale because the systems are not AI-ready. The heat-map tells that story in its color pattern faster than any paragraph.

Two design choices make the heat-map land. First, never show the heat-map without the evidence notes one click away, because a leader who acts on color without evidence will make a bad call and blame your instrument. The heat-map is the headline; the notes are the article. Second, show the prior quarter's heat-map beside the current one the moment you have one, because the delta - "design systems moved from red to amber this quarter" - is the single most persuasive thing you can put in front of a leader who is wondering whether the AI investment is working. A maturity assessment that is run once is a snapshot; run quarterly it becomes a trend line, and trend lines are how you defend a budget.

From Heat-Map to a 90-Day Priority List

A heat-map that does not produce action is design theater. The other half of the artifact, and the half that proves you are a strategist rather than an auditor, is the 90-day priority list: the small number of moves that will most improve the team's maturity in one quarter, chosen deliberately, with the reasoning visible. The discipline here is ruthless prioritization. You will be tempted to attack every red cell at once; resist it, because a team that tries to fix six dimensions in a quarter fixes none and exhausts itself in the attempt.

Prioritize by leverage, not by redness. The reddest dimension is not always the one to fix first; the one to fix first is the one where a quarter of effort unlocks the most downstream value. In 2026 that is very often design systems, because an AI-readable token-and-component layer is the thing that makes every other AI workflow stop producing drift, which means fixing it raises the effective maturity of tools, workflows, and brand systems at once. A single high-leverage move can light up several cells indirectly, which is exactly the kind of reasoning a leader wants to see: not "we will work on everything," but "we will do this one thing because it makes four other things work."

What a Good Priority Item Looks Like

Each item on the 90-day list has four parts, or it is a wish rather than a plan: the dimension it moves, the specific move ("tokenize the semantic color layer and publish it in DTCG format with CI validation"), the named owner, and the observable signal that it worked ("generated components in the pilot read from real tokens; off-system color in new work drops to near zero"). Three to five items is the right number for a quarter. More than five and you are not prioritizing; fewer than three and you are under-using a full quarter of a team's capacity. Each item should also name the dimension it moves from which level to which level, so the next assessment can confirm the move actually happened rather than just felt like it did.

The priority list is also where you connect this lesson forward to the rest of L4. The 90-day list feeds directly into the use-case inventory and the quarterly initiative pattern; the maturity gaps you name here become the use cases you score there and the initiatives you sequence after that. The assessment is the diagnosis; the roadmap is the treatment plan. Doing the diagnosis first is what keeps the treatment from being a list of tools someone read about on LinkedIn.

A Worked Example: Scoring a Real-Shaped Team

Make it concrete. Imagine a ten-person product-design team at a Series C company. You run the assessment and here is what the evidence shows. People: a bimodal distribution - three ICs are fluent and verifying their AI work, the other seven range from cautious to frozen, and one senior IC is openly resistant. Median capability is low and the distribution is two-speed, so you score People at Level 2 with the note "fluency concentrated in 3 of 10; two-speed risk." Tools: lots of licenses, including overlapping image models and three prototyping tools with no decision rule, plus shadow ChatGPT spend - tool-rich, governance-poor, so Level 2, note "no decision rule; overlapping spend."

Workflows: exactly one documented workflow (research synthesis), used by four people; everything else is osmosis - Level 2, note "one of N workflows documented." Design Systems: a gorgeous Figma library, zero machine-readable token layer, no MCP-readable component docs - this is the trap dimension, beautiful to humans and invisible to agents, so Level 1, note "no DTCG tokens; agents cannot read the system." Brand Systems: brand lives in two senior heads, no anchor set, recent launch assets drifted generic - Level 1, note "no brand-anchor set; drift observed." Design Ops: no provenance log, no IP policy beyond instinct, no verification-tax accounting - Level 1, note "no provenance log; IP exposure unmanaged."

The heat-map is mostly red and amber, with the two systems dimensions and design ops in the red. The story writes itself: capable-enough people with too many ungoverned tools, producing work that will drift at scale because neither the design system nor the brand system is machine-respectable, with no governance to catch the consequences. The 90-day list does not try to fix all of that. It names three moves: tokenize and DTCG-validate the semantic layer (Design Systems, Level 1 to Level 2, owner the systems lead) because it is the highest-leverage unlock; stand up a provenance log and a one-page IP policy (Design Ops, Level 1 to Level 3, owner you) because it is cheap and closes a real liability; and write a tool decision rule that retires the overlapping licenses (Tools, Level 2 to Level 3, owner a senior IC) because it is fast and saves money the same quarter. People, brand, and the deeper workflow work are named explicitly as next quarter, so the leader sees a sequenced plan, not an omission.

Defending the Assessment Upward

The last move is presenting it, because an assessment that stays in your drawer changes nothing. When you take the heat-map up, lead with the shape and the one-sentence story, not the methodology - executives want the diagnosis and the plan, not the scoring rubric, which lives in an appendix for when someone challenges a score. Frame the red cells as managed risks with owners and dates, not as confessions, because the same fact ("our design system is not AI-readable") reads as a liability when stated as a confession and as competent risk management when stated as "this is our highest-leverage gap and here is the 90-day plan to close it." The facts are identical; the framing determines whether the conversation funds you or cuts you.

Expect two challenges and have answers ready. "Why are we so red?" - because this is an honest baseline, every team running an honest assessment in 2026 is red somewhere, and a team claiming all green is not measuring honestly; the value is that we can now show movement. "How do we know the plan will work?" - because each priority item has an observable signal we will re-measure next quarter, so this is not a promise, it is a hypothesis with a test attached. That second answer is the one that earns you the resources, because it tells the leader you are running the team like a system that learns, which is exactly the posture that survives the cycle.

Key Takeaways

• A feeling ("we're behind on AI") cannot be funded, defended, or compared across quarters. The maturity assessment is the translation layer that renders what you already know about your team in the structured, scored form the org respects, turning a worry you carry alone into an artifact a CFO can act on.
• Audit six dimensions, because each fails differently and is owned by a different lever: People, Tools, Workflows, Design Systems, Brand Systems, and Design Ops. Skip one and it blindsides you within a quarter. Score people on distribution, not average, to catch the two-speed-team trap.
• Score every dimension on the same four-level scale - Ad Hoc, Emerging, Defined, Optimized - so six things fit on one heat-map and quarters become comparable. Score the median person and the documented practice, never the best designer on their best day; a practice that lives in one head is Level 2 at most.
• Render the result as a heat-map (red to deep green across the six rows) so a leader reads the shape in ten seconds, but never show it without the one-line evidence notes, and show the prior quarter beside it the moment you have one - the delta is the most persuasive thing you can present.
• Convert the heat-map into a 90-day priority list of three to five items, each with a dimension, a specific move, a named owner, and an observable success signal. Prioritize by leverage, not redness - in 2026 the AI-readable design system is often the highest-leverage fix because it raises several other dimensions at once.
• The assessment is the diagnosis that feeds the rest of L4: its named gaps become the use cases you score and the initiatives you sequence. Run it quarterly so the snapshot becomes a trend line, because trend lines are how you defend a budget.
• Present it as managed risk with owners and dates, not confession. The same fact reads as liability or as competent risk management depending on the framing, and "each item has a signal we will re-measure" is the answer that earns resources because it shows you run the team as a system that learns.