You bought Claude Teams or Claude Enterprise. Three people use it heavily. The rest of the team barely touches it. That sentence is in our intake notes more often than any other. Sometimes the number is two, sometimes it's five, but the shape of the problem is identical: licensed seats outnumber active users by a factor of three or four, and nobody at the company can produce a defensible adoption number when leadership asks.
This is not a Claude problem. The platform works. It's an installation problem. Buying a license is the easy part. Turning a license into a measurable adoption story by a named board or LP date is the actual work, and there's a sequence to it that produces results in 60 days. Here it is.
What "measurable ROI" actually means
Before the playbook, the definition. "Measurable ROI" in this context is not a single number. It is three numbers, one per axis, all anchored to a baseline captured before training started:
- Adoption number. Behavior. From platform telemetry — active users, project usage, Skill invocations per named workflow. Pulled off the Claude Enterprise Analytics dashboard or API. Objective, automated, content-free.
- Enablement number. Competence. Maturity-scale delta per person across prompting, Project setup, Skill use, and output review. From forms — pre/post.
- ROI number. Value. Self-reported hours recovered per named workflow, dollarized at a loaded rate the client provides. Plus the validator Skill's quality-flag rate, which is the one ROI signal the system measures directly.
Pair the three and the dashboard survives a budget review. Skip one and there's always a "but" that leadership can poke at. The methodology behind this lives at How to Measure AI Adoption Without Making It Up.
The five-stage cadence
The playbook below applies whether the engagement runs 60 days or 90. The stages don't change. The cadence and depth scale. Here is the full sequence.
Stage 1 — Intake (Day 0)
Everything that has to happen before training starts. Skip any of it and the engagement starts with a gap.
- Skills assessment + department-head intakes. Every participant fills out a maturity self-rating. Department heads do a 30 to 60 minute intake call to surface the team's actual tools, workflows, and time sinks.
- Baseline metrics. Platform baseline (snapshot current Claude Enterprise usage), enablement baseline (the maturity ratings above), workflow baseline (each person picks one target workflow and records time, frequency, dollar value).
- Access setup. Primary Owner issues the Analytics API key with
read:analyticsscope, or commits to CSV exports on a cadence. Build seats provisioned for the consultant. Participant emails collected so the post-engagement re-measure can be matched to the baseline.
The constraint that breaks late starts: the Claude Enterprise Analytics data window only goes back to January 1, 2026, and only runs 90 days. The baseline has to be captured before Session 01, never after. If the engagement starts and the baseline doesn't, the window can move past the start date and the floor disappears. This is the single most important sequencing constraint in any Claude rollout.
Stage 2 — Strategy and scoping
Before training kicks off, three decisions get locked.
- Which workflows become Skills. Three named workflows, picked for being real, painful, and high-frequency. This is the decision that makes the Adoption axis countable later. Skill-invocation count equals workflow-execution count. Without this decision, adoption stays qualitative forever.
- Target metrics per axis. Specific numbers, not aspirations. "We want X active users by Day 60. We want Y average maturity rating across the team. We want Z hours recovered per analyst per week."
- The autonomy ladder per workflow. Where on the assist-vs-autonomous gradient each workflow lives at end of engagement. Some workflows want a human checking output every time. Some are fully autonomous behind a validator. Pick per workflow.
Stage 3 — Training series (Phase 1)
The classroom phase. Seven 90-minute sessions across departments, sequenced so foundations land first and role-specific tracks land in week two and three.
The thing to measure during Stage 3 is not attendance — that's the wrong metric and always has been. The thing to measure is participation depth: who builds versus who observes. This is the enablement signal you can only catch in the room. A senior analyst who builds during the session is on a different trajectory than a senior analyst who observes. Track the ratio and the dashboard's enablement axis populates itself.
Stage 3 also runs the early-engagement check on platform telemetry. If a deployed Skill is going dormant in Week 2, that's the time to surface it, not Week 6.
Stage 4 — Workflow builds (Phase 2)
Runs in parallel with Stage 3, not after it. Three named workflows built as Skills and deployed to the client's AI Portal. Plus a validator Skill that runs over output and logs quality flags — the one ROI signal the system measures directly.
Stage 4 is where the rollout converts from "training that fades" into "infrastructure the team has to walk past every day." Once three Skills are deployed and the team is invoking them, the team's daily work has Claude in it whether anyone reminded them to use it that day or not. That is the difference between adoption that takes and adoption that fades.
Stage 4 also produces a written data security protocol — the answer to "what data goes into Claude, what doesn't, who has admin rights, how shared Projects are managed." This is non-negotiable for regulated industries (PE, family offices, healthcare, legal, insurance) and load-bearing for any team where compliance is going to look at the rollout.
Stage 5 — Re-measure and handoff (Day 60)
The dashboard work. Three deltas, one document.
- Platform delta. Active users, Project usage, Skill invocations versus Day 0 baseline. Pulled off the Analytics API.
- Enablement delta. Maturity-scale movement per person. Re-measure form sent to the same people who filled out the baseline, matched by email.
- ROI. Hours recovered per named workflow, dollarized at the loaded rate the client provided at intake. Plus the validator's quality-flag rate.
Output: a wrap-up brief usable as a board or LP readout. Five minutes to read. The whole engagement defensible on one document.
What the dashboard looks like in practice
The anchor engagement for this playbook in Q2 2026 was a 25-person capital management firm. They had Claude Teams licensed. They had four power users and roughly twenty seats that logged in once a week or less. They had a hard September LP meeting where leadership wanted to point to AI adoption as a value-creation story.
The Day 0 baseline: four weekly active users, average maturity rating 2.1 across the team, zero deployed Skills, no validator. Day 60 dashboard: 19 weekly active users, average maturity rating 3.4, three deployed Skills (IC memo generator, board pack assembler, research validator) with 47, 21, and 96 invocations in the last 30 days, ~6.8 hours saved per analyst per week at the team's loaded rate. The wrap-up brief was three pages. The LP meeting got a defensible adoption story.
The platform did most of the measurement work. The forms filled in the value side. The whole thing was provable because the baseline was captured before Session 01 and the three named workflows were built as Skills so the platform could count them.
What you do if you're already mid-rollout
If you're three or six months into a Claude rollout that's fading, you don't need to start over. You need to retrofit the missing pieces. In order of leverage:
- Capture whatever baseline you can right now. Even an imperfect baseline is better than no baseline. Pull whatever's available off the Analytics dashboard, send a maturity self-rating form to the team this week, ask department heads to identify their three highest-leverage workflows.
- Build three Skills around the workflows that surfaced. The build is the hardest gap to retrofit and the highest-leverage to close. Three deployed Skills changes the shape of the next 30 days.
- Issue the Analytics API key. One Primary Owner action. From that point forward, the platform measures behavior automatically.
- Schedule a 30-day re-measure. Send the same form again. Pull the platform telemetry delta. Now there's a directional dashboard, even if the starting point was retroactive.
This is the salvage path. It works. It is more expensive than starting with the structure in place because some of the data is fuzzier than it would have been from Day 0. But it is the difference between a rollout that survives the next budget review and one that doesn't.
The 60-day vs. 90-day decision
The five-stage cadence above runs at either pace. The differences come down to scope and the post-engagement window.
The 60-day variant is built for the team that needs a story to tell by a moving date — a September LP meeting, a Q4 board, a leadership review with a calendar slot. Phase 1 is seven 90-minute sessions. Phase 2 is three named workflow builds plus a blueprint of additional automations to build next. Async support runs across the engagement plus 14 days after. Single re-measure at Day 60.
The 90-day variant is built for the team that needs the story plus one additional automation live before the engagement closes. Same Phase 1 seven sessions, plus two follow-up working sessions. Same three named workflow builds plus blueprint, plus one build from the blueprint delivered inside the engagement. Async support runs across plus 30 days after. Two re-measures (Day 60 and Day 90) so the dashboard has a trendline, not just a delta.
Most teams in the 5-to-30-person range run the 60-day variant. Teams in the 30-to-50-plus range or teams with one specific automation that needs to land in-engagement run the 90-day.
What this is and what it isn't
This is an operating-layer install. Training installed across every department, three Claude-native workflow builds delivered, an admin and governance layer, and a measurable adoption dashboard. It is priced on outcomes, not on training delivery.
It is not a workshop. It is not a strategy document. It is not a free discovery call. The methodology lives at How to Measure AI Adoption Without Making It Up and the structural gaps it closes are walked through at The Three Things Every AI Rollout Misses. The training-tier comparison is at Why Most AI Training Fails.
If your team has Claude Teams or Enterprise licensed and the adoption story isn't where it needs to be by a named date, this is the playbook that produces the story.
We productized this playbook. The 60-Day Claude Rollout™ — from $22,500 for 5 to 15 people, $32,500 for 15 to 30, $48,500 for 30 to 50-plus. 90-day variants priced separately.
Nicole Patten is the founder of Elevate Online and runs a Claude-specific training practice. She spent 7 years at Google as a Senior UX Engineer before dedicating her career to helping teams use AI responsibly and effectively. 100% of her business runs on Claude.