Scaling Claude Cowork from pilot to organization

TL;DR. Take a successful 6-week pilot in one function and stand up a credible Wave 2 covering 2–3 more functions, then a steady-state model. Wave 2 is 2–3 functions in parallel; never one, never five. By month four you need a named enablement owner — even part-time. The companies that don't name one are still running Cowork in a single function a year later.

The two scale models#

Two valid approaches, not interchangeable:

• Replication — repeat the same function pattern in adjacent business units (e.g., Finance close in three regions).
• Expansion — apply the method to a new function (e.g., Finance → Marketing Ops → Legal Ops).

Most mid-market companies need both, sequenced. Replication is faster but limited; expansion is slower but unlocks bigger ground.

The Wave 2 sizing#

• 2–3 functions in parallel (not 1, not 5).
• Each with its own pilot owner.
• Shared sponsor — the original Wave 1 sponsor, now experienced.
• A Tinkso engineer or internal lead floating across all three.

The temptation to expand to five at once is strong; the math against it is straightforward — focus dilutes, attention drops, nothing ships.

What gets reused from Wave 1#

• The workspace pattern.
• The CLAUDE.md skeleton, customised per function.
• The 6-week rollout shape.
• The shadow + pain/gain templates.
• The two-metric ROI discipline.

The hardest reuse: the change-management practices. They look optional in Wave 2 because the org "already knows Cowork." They are not. Skip them and Wave 2 underperforms Wave 1 — which is the worst political signal you can send.

What you build new in Wave 2#

• Function-specific workflows (won't transfer).
• Function-specific MCP connectors (different SaaS per function).
• Function-specific skills.
• A new pilot owner per function.

Each new function still gets its own Observe beat. Compressed, but not skipped.

The steady-state organisation#

Once three or more functions are live:

• A Cowork enablement function — one or two people, often part of IT or RevOps.
• A monthly governance forum — see Rollout governance.
• A library of skills owned by the enablement team.
• A self-serve onboarding pack for new operators.
• Quarterly reviews of seat mix and connector inventory.

Without these, Cowork survives but does not compound. With them, value compounds across functions and the per-function rollout cost halves and halves again.

The enablement function job description#

A short, real description for the role spec:

• Maintain workspace patterns and CLAUDE.md skeletons.
• Run the monthly governance forum.
• Own the skills library.
• Onboard new operators in a one-day session.
• Liaise with security, IT, and Anthropic.
• Surface three cross-functional opportunities per quarter.

This role pays for itself by month six in any deployment past 50 active seats.

When to bring in outside help#

• A new connector you don't have engineering capacity to build.
• A regulated function where compliance documentation is heavy.
• A function whose workflow you cannot crisp without external eyes.

For mid-market, outside help usually means Tinkso for 2–4 week sprints, not a six-month consulting engagement.

The 12-month horizon#

Realistic milestones for a mid-market company starting from one Wave 1 pilot:

• Months 1–2 — Wave 1 pilot ships in one function.
• Months 3–4 — Wave 2 covers 2–3 functions.
• Months 5–6 — enablement function in place; first cross-functional skills.
• Months 7–9 — steady-state operations; quarterly governance live.
• Months 10–12 — skills library matures; ROI compounds across functions.

This assumes consistent sponsorship and no major reorgs. Both are heroic assumptions in mid-market; if either changes, expect a quarter of slippage.

Tinkso's take#

Scaling is not a project. It is a function. The mid-market companies that get value compounding past month six are the ones that named a Cowork enablement owner — even part-time — by month four. The ones that did not are still running Cowork in a single function a year later, wondering why.

Try this#

At your week-6 ROI checkpoint, ask the sponsor one question: "Who owns Cowork enablement on day one of next quarter?" If the answer is not a name, you have a Wave 2 problem to solve before you have a Wave 2 plan to make.