AI Adoption Level Guide
This adoption canvas sets the success path for change-management — how people progress with AI and compound their systems.
Leverage × efficiency × effectiveness — how much real value each person produces with AI.
What people ship, how they climb, how the team compounds.
Two people at the same level can produce wildly different outcomes — a focused & prioritized L1 can out-perform a flashy L4, but a prioritized L4 is unstoppable. We want to climb the rubric together — empower people to create a bigger impact than they could on their own.
This matrix is inspired by Zapier’s AI Competency rubric, which measures an individual's impact with AI for hiring standards & performance reviews. Each dimension of impact (Unacceptable → Aware → Capable → Adoptive → Transformative) is mapped to the level guide.
How to read this: Rows ladder Zapier’s competence titles from highest impact down to the anti-pattern. The Zapier-style template column is the self-assessment statement you’d write at that tier — read just this column for the simple overview if L0–L5 is new to you. L0–L5 columns show what that tier looks like at each skill level.
| Impact Dimension (Zapier) | ZAPIER-STYLE TEMPLATE Self-assessment statement at this tier | 00 PRE-ADOPTER | 01 ONE-OFF AI | 02 CONNECTED AI | 03 AUGMENTED AI | 04 AUTONOMOUS AI | 05 AI NATIVE |
|---|---|---|---|---|---|---|---|
🏆 Transformative Multiply the Team makes others more effective | “I led [initiative] using [AI tools] that delivered [quantified outcome] for [team/org]. I learned [lesson] and addressed [risk/ethic].” | “I haven’t pushed AI on anyone yet — I want to understand it before I tell others what to do.” | “I showed a teammate the prompt I use for cleaning up meeting notes; she now uses it after every standup and saves ~20 min/day.” | “I wrote up my CLAUDE.md + workflow in a Loom and README; two teammates copied it and one extended it with their own role context.” | “I ran a 30-min team session on our shared prompt library and MCP setup — 4 of 6 teammates are now using it weekly, tracked in a usage sheet.” | “I built an agent the team triggers in Slack to draft client follow-ups; it handles ~40/week and the team only reviews.” | “I redesigned our hiring rubric and role descriptions around AI-native work — two new hires ramped in 3 weeks instead of 8, and we retired one role that had become fully automatable.” |
Adoptive Build with AI creates reusable AI assets | “I introduced [AI tool] across [scope]; we now [outcome] with [Δ time / Δ quality]. Approach: [method].” | — | “I keep a Notes file with 3 prompts I’ve tuned over a month — email tone, meeting digests, QBR prep — and reuse them weekly.” | “I committed a CLAUDE.md with my role, goals, voice, and active projects; every session loads it and outputs land closer to final on the first pass.” | “I wired an MCP to our CRM and shipped a shared prompt the team uses to pull account context before client calls; 3 teammates use it weekly.” | “I built a custom agent that ingests every weekly status report, flags risks, and posts a synthesis to our team channel each Monday — no human in the loop.” | “I run 4 scheduled agents with defined handoffs (intake → triage → draft → review queue); I monitor a dashboard rather than write outputs.” |
Capable Use AI in Work applies AI to your own output | “I use [AI tool] to [task]; it changed how I [outcome] and saves me ~[time]/week.” | “I’ve watched a few videos on AI use cases for my role but haven’t opened a chat tool yet.” | “I draft first versions of emails, summaries, and research in Claude before editing — time-to-first-draft on a 5-bullet exec update dropped from ~30 min to ~8 min.” | “I work inside Claude Code with a persistent CLAUDE.md, so I no longer re-explain my role each session — outputs feel like ‘me’ from prompt one.” | “My start point is an AI-native surface (Claude Code / CoWork / IDE); I bounce to a browser only for sourcing, not authoring.” | “I delegate defined sub-tasks (competitor research, PRD draft sections, 90-min meeting summaries) to agents and spend my time reviewing rather than producing.” | “Most of my day is reviewing and steering agent output — authoring is the exception, not the default.” |
Aware Spot Opportunities notices where AI applies | “I’ve noticed [recurring task / pain]; it feels like AI could help with [hypothesis] — here’s where I’d start.” | “I noticed our weekly client check-in produces the same 4 sections every time and wrote ‘this seems automatable?’ in a Notion doc.” | “I picked one already-on-my-plate task (drafting customer follow-ups) and tried Claude on it; result beat my alone-draft, so I kept doing it.” | “I mapped the parts of my role with repeatable structure — status updates, customer research, meeting prep — and started a tuned prompt for each.” | “I spotted that 5 PMs were each writing weekly status reports from scratch and proposed one shared prompt + CLAUDE.md the team adopted.” | “I traced our quarterly planning workflow end-to-end (intake → synthesis → write-up → review) and identified the synthesis step as the choke point an agent could own.” | “I led a re-scope of what my team actually does — we kept the strategy and stakeholder work, retired the artifact-production work, and the headcount mix shifted.” |
Unacceptable When it spins wheels skill present, impact missing | “I built / learned / adopted [thing], but [no one uses it / nothing changed / I can’t point to an outcome].” | “I spend more time arguing about AI in meetings than I would spend trying it on one real task.” | “I run every prompt through Claude, ChatGPT, and Gemini to pick ‘the best’ — net slower than if I’d just used one and shipped.” | “I have an immaculate CLAUDE.md I haven’t opened in 60 days, and outputs that still don’t feel like mine.” | “I built a ‘team prompt library’ that’s a folder of dead links and stale screenshots — no one references it.” | “I shipped a beautiful agent that solves a problem no one had; it runs weekly and is opened by no one but me.” | “I operate a fleet of agents producing reports no human downstream consumes — efficient at producing waste.” |
This isn’t new with AI. Every wave of ops, automation, and productivity tooling produced the same pattern:
- RPA, 2018: bots automating processes the business had already changed.
- Slackification, 2015: channels piping every CI build, JIRA update, and PR comment — read by no one.
- Dashboards, forever: the 47-chart executive dashboard opened weekly only by its builder.
- Workflow tools, today: Zapier zaps firing 4,000×/month moving data nobody queries.
AI doesn’t change the rule. It just makes the bottom row cheaper to produce — which means the gap between motion and impact widens unless you watch for it.
- Read the Zapier-template column first if L0–L5 is new — pick the tier whose statement you could honestly fill in today.
- Find your honest baseline level. The level where the Transformative row feels earned, not aspirational, is your real level.
- Check the Unacceptable row at your level. If any of it sounds familiar, that’s the leak to fix before climbing.
- Aim one tier up on one row — not the whole column. Pick the dimension where moving up would change a real outcome this month.
An L1 with three reused prompts and a teammate who adopted one is producing more impact than an L4 whose agent nobody runs. Climb the rubric. But cash the impact.
Leverage × efficiency × effectiveness — how much real value each person produces with AI.
What people ship, how they climb, how the team compounds.
Two people at the same level can produce wildly different outcomes — a focused & prioritized L1 can out-perform a flashy L4, but a prioritized L4 is unstoppable. We want to climb the rubric together — empower people to create a bigger impact than they could on their own.
This matrix is inspired by Zapier’s AI Competency rubric, which measures an individual's impact with AI for hiring standards & performance reviews. Each dimension of impact (Unacceptable → Aware → Capable → Adoptive → Transformative) is mapped to the level guide.
How to read this: Rows ladder Zapier’s competence titles from highest impact down to the anti-pattern. The Zapier-style template column is the self-assessment statement you’d write at that tier — read just this column for the simple overview if L0–L5 is new to you. L0–L5 columns show what that tier looks like at each skill level.
makes others more effective
Template: “I led [initiative] using [AI tools] that delivered [quantified outcome] for [team/org]. I learned [lesson] and addressed [risk/ethic].”
“I showed a teammate the prompt I use for cleaning up meeting notes; she now uses it after every standup and saves ~20 min/day.”
creates reusable AI assets
Template: “I introduced [AI tool] across [scope]; we now [outcome] with [Δ time / Δ quality]. Approach: [method].”
“I keep a Notes file with 3 prompts I’ve tuned over a month — email tone, meeting digests, QBR prep — and reuse them weekly.”
applies AI to your own output
Template: “I use [AI tool] to [task]; it changed how I [outcome] and saves me ~[time]/week.”
“I draft first versions of emails, summaries, and research in Claude before editing — time-to-first-draft on a 5-bullet exec update dropped from ~30 min to ~8 min.”
notices where AI applies
Template: “I’ve noticed [recurring task / pain]; it feels like AI could help with [hypothesis] — here’s where I’d start.”
“I picked one already-on-my-plate task (drafting customer follow-ups) and tried Claude on it; result beat my alone-draft, so I kept doing it.”
skill present, impact missing
Template: “I built / learned / adopted [thing], but [no one uses it / nothing changed / I can’t point to an outcome].”
“I run every prompt through Claude, ChatGPT, and Gemini to pick ‘the best’ — net slower than if I’d just used one and shipped.”
This isn’t new with AI. Every wave of ops, automation, and productivity tooling produced the same pattern:
- RPA, 2018: bots automating processes the business had already changed.
- Slackification, 2015: channels piping every CI build, JIRA update, and PR comment — read by no one.
- Dashboards, forever: the 47-chart executive dashboard opened weekly only by its builder.
- Workflow tools, today: Zapier zaps firing 4,000×/month moving data nobody queries.
AI doesn’t change the rule. It just makes the bottom row cheaper to produce — which means the gap between motion and impact widens unless you watch for it.
- Read the Zapier-template column first if L0–L5 is new — pick the tier whose statement you could honestly fill in today.
- Find your honest baseline level. The level where the Transformative row feels earned, not aspirational, is your real level.
- Check the Unacceptable row at your level. If any of it sounds familiar, that’s the leak to fix before climbing.
- Aim one tier up on one row — not the whole column. Pick the dimension where moving up would change a real outcome this month.
An L1 with three reused prompts and a teammate who adopted one is producing more impact than an L4 whose agent nobody runs. Climb the rubric. But cash the impact.