Change management — making Claude Cowork stick past week 8 — The Cowork Bible — Tinkso

TL;DR. Not a Prosci ADKAR diagram. A small set of practices that make a Cowork rollout still in active use eight weeks after the pilot ends. Most rollouts that look healthy at week six die between week eight and week sixteen because nobody is doing the small communication work.

The mid-market reality#

Most mid-market companies do not have a change-management function. The pilot owner is the change manager whether they signed up for it or not. This page is the cheat sheet that makes that role doable in 30 minutes a week, not 30 hours.

The five practices#

Make the win visible. Publish a one-page what we shipped this week inside the team. Two paragraphs, one number. Weekly during the pilot, monthly after.

Name the new normal. "We use Cowork for proposals on Mondays." Specific. Time-bound. Not "we use AI."

Pair-write CLAUDE.md. When a new prompt is added, two operators write it together. The second operator's questions surface the assumptions the first operator did not realise they were making. Forces transfer.

Surface failures. Weekly 15-minute what didn't work round. Lower the bar to share a Cowork misfire — those misfires are the fastest path to a sturdier prompt.

Celebrate quietly. A Slack emoji reaction for "this artifact came out of Cowork" is enough. No company-wide announcements. The point is recognition, not theatre.

What kills adoption#

A sponsor who goes quiet after week two.
A pilot owner who switches projects.
A change in priorities that pulls the operator off Cowork for three weeks (loss of habit).
A bad Cowork output that wasn't caught (loss of trust).
A comp model that pays for the old way of working.

The last one is the most insidious. If the bonus is tied to the old metric, the team will go back to the old method even when the new one is better.

Communication cadence#

When	Who	Format	Length
Weekly during pilot	Pilot owner → sponsor	Email	5 bullets
Bi-weekly during pilot	Pilot team	Slack stand-up	15 min
Week 6	Pilot owner + sponsor + Tinkso	Review meeting	60 min
Month 2 onward	Pilot owner → sponsor	Monthly note	1 page
Quarterly	Sponsor → exec team	Update	1 page + 1 chart

That is the entire schedule. Stick to it.

The "new normal" check at week 8#

Two weeks after the pilot ends, the sponsor asks the operators three questions:

Are you still using Cowork in your daily work?
What part has stuck? What hasn't?
What would make the not-stuck parts stick?

The answers tell you whether to scale or stabilise.

Scaling to new teams#

Do not copy-paste the pilot's prompts. Run a mini-Observe beat for the new team.
Reuse the workspace pattern, CLAUDE.md skeleton, and the pilot owner as a coach.
Expect Wave 2 to take half the time of Wave 1, not zero.

The biggest scaling mistake is assuming Wave 2 is free. It is cheaper, not free.

The most common late-stage failure#

At month four, the pilot team has stopped iterating. Cowork is done. This is the moment when adoption decays — small bugs accumulate, prompts go stale, the team stops sharing wins. Counter with a quarterly governance review (see Rollout governance) and a refreshed pain/gain map.

Tinkso's take#

Change management is unglamorous and decisive. Most rollouts that look healthy at week six die between week eight and week sixteen because nobody is doing the small communication work. We bake a 30-minute monthly "still working?" check into our retainer for the first six months, and it pays for itself in continued adoption.

Try this#

Schedule the week-8 check now, while you're reading this. Two weeks after pilot end. Thirty minutes. Three operators. The forcing function is the calendar entry; the conversation is what makes it work.