12 · Anthropic's adoption frameworks

Anthropic publishes its own maturity model, rollout roadmap, use-case selection criteria and governance postures — mostly in places its learning hub never links to. These are the most directly liftable assets found anywhere in the official material.

The five-level maturity model

This is the single best asset in Anthropic's published material, and it does not appear in the product documentation at all. It gives an executive audience a ladder rather than a binary.

The framing line — the best positioning sentence Anthropic has published

"Chat is everyone's Level 0. Claude Cowork is where Levels 1 through 4 happen for knowledge work, and Code is where the same progression happens for engineering."

This is a cleaner articulation of the same-engine-two-products point than anything in the product documentation, and it maps directly onto §02. It also reframes the extensibility ladder in §06 — instructions → skills → plugins — as an organisational progression rather than a technical one.

And the expectation-setting line that goes with it

"The job of a deployment is to get every user one level higher than they are now and to make sure the path upward is paved; nobody needs to reach Level 4 in the first month."

Chat vs Cowork vs Code — Anthropic's own surface-selection table

Published in the same guide. It independently corroborates the decision logic in §10, which is worth knowing: you are not extrapolating, you are agreeing with them.

"The three share the same Claude underneath; what changes is the workspace around it."

Choosing a first use case — four criteria

Verbatim from the guide's "Choosing your first use case" section. Note that the fourth is doing real work — it is the one that argues for encoding tacit expertise rather than automating obvious drudgery.

The line that reframes the whole value proposition

"When your best analyst's workflow lives in a skill rather than in her head, it stops being tribal knowledge and becomes organizational infrastructure."

And the accompanying claim about where the return actually shows up:

"This is where functional leads tend to see the biggest return—not doing the same work faster, but doing the work that mattered but was never a P0."

The six-month adoption roadmap

Compare this against the staged path I proposed in §10. They agree on shape — champions first, encode second, distribute third — but Anthropic's is more concrete on timing and mine adds the instrumentation step. Use whichever suits the room; say which one you're showing.

Starting points and metrics by function

Measuring adoption — the three-question model

From the Cowork Enterprise Administrator Guide:

Note the dependency this creates: on Cowork, questions one and two require the OpenTelemetry pipeline described in §08 — which is Team/Enterprise only. On Claude Code, the analytics dashboard covers it natively. If you intend to measure a Cowork rollout, the telemetry has to be stood up before the pilot starts, not after.

Three governance postures

From the Scaling Workflows playbook. This is the cleanest published answer to "how much should we let people build?"

Accompanying numbers from the same source: one champion per 25–50 users, and a Phase 1 target of over 70% weekly active users by week 4.

The risk framing Anthropic uses — and it's a good one

"When individuals adopt AI tools without oversight you get shadow AI: dozens of private workflows nobody can see, audit, or improve."

Deploying Claude Across Your Organization, p.11 — not the Scaling Workflows playbook quoted above it.

Pair this with the Cowork-specific fact from §08: Cowork projects are not shareable and local session data is not centrally managed. The product's own defaults push toward exactly the shadow-AI state this warns about. Naming that tension is the most useful thing you can do for a risk stakeholder.

The adoption argument, in Anthropic's words

"Deploying the technology is the easy part; getting an organization to actually use it is much harder."

"Define success before you start among relevant stakeholders. 'Hours saved per week' is measurable. 'Transformation' is not."

"The pattern is bottom-up discovery, top-down scale."

"You don't need every employee using the same plugins to call a deployment successful. You need every employee more productive than they are today, and a clear path to the next rung of the adoption ladder."

Claude Code's parallel rollout model

From Scaling Agentic Coding Across Your Organization. Structurally the same shape as the Cowork roadmap, with an engineering flavour:

1. Start with super users — a pilot group of 20–50 developers already using AI tools.
2. Launch with a hackathon — an org-wide kickoff event.
3. Scale through internal expertise — the pilot group becomes the advisory body.

Proposed ROI metrics: sprint throughput · task completion time · migration velocity · developer satisfaction · onboarding duration · cross-functional efficiency.

"Agentic coding shifts software development from writing every line to guiding implementation."

Named customer outcomes

Useful because they are attributable and quantified. Use sparingly — one or two, chosen for resemblance to the room.

Quotable customer lines

"Claude Cowork helps teams do work at a scale that was hard to justify before. The human role becomes validation, refinement, and decision-making. Not repetitive rework."
— Joel Hron, CTO, Thomson Reuters

"What would have required a team of engineers building a custom React app, Claude Cowork delivered in 45 minutes. And it's more adaptive than anything we would have built."
— Matt Benyo, Director of AI Initiatives, Jamf

"The barrier between 'having an idea' and 'shipping something' has collapsed."
— Larisa Cavallaro, Zapier

Two more assets worth knowing about

The Claude Code communications kit — drop-in change management

A complete rollout comms package: launch announcement with audience variants, a 20-message drip campaign for Slack or Teams, and a quick-reference FAQ. Its pre-launch checklist includes "Recruit C-suite sponsor (exec-sent launches show higher adoption)," and there is a dedicated Executive Sponsor Send variant.

"The announcements that actually drive adoption are the ones that read like someone at your company wrote them."

Anthropic's own caveat on the kit: it is "draft copy, not finished copy."

There is no Cowork equivalent. Which is an odd gap, given Cowork is the product a non-engineering rollout actually needs comms for. The Claude Code kit adapts with moderate effort.

The open-source plugin library

Anthropic publishes plugins covering "sales, legal, finance, marketing, product, HR, and more," and has open-sourced its own Legal plugin. These make excellent lab exercises: install one, read the skill files, then modify one for the participant's own team. It turns the abstract Level 2→4 progression into something they touch.