Claude Enterprise now treats chat and Cowork as one thing. Pro and Max users got the merged version in September, and Enterprise organizations have had it since October. A conversation starts as a normal chat. When the task needs files or code, Claude continues in a cloud environment that Anthropic manages, and it keeps going after the person closes their laptop.
The old choice was also a cost control
Until now an employee had to choose: chat for the quick question, Cowork for the bigger job. Choosing Cowork was a small act of judgment, a moment of "I am about to hand something over." On Enterprise, the seat fee covers access only and every token is billed separately at standard API rates, so that moment was also the spending decision.
With the two merged, Claude makes the call. It will usually make it well, because a job that needs files and code really does need them. The change is that nobody on the team experiences it as spending anymore.
Scheduling turns one job into a standing cost
The Monday report shows how this adds up. A team lead asks Claude to read the week's shift logs and stock files, flag what slipped and build five slides. The result arrives as a document and slides, as we covered when Docs and Slides came to chat. Then the lead schedules it for every Monday, and it runs with no laptop open.
Say each run reads 2 million tokens and writes 100,000. At Opus 5.5 prices of $4 per million tokens read and $20 per million written, that is $8 plus $2, so $10 a run. Fifty-two Mondays cost $520 for one report. Forty team leads doing the same cost about $21,000 a year.
That is still a good trade. If each report saves its lead two hours a week, the 40 leads get back more than 4,000 hours, so each hour costs about $5. But nobody approved a $21,000 line. Forty people each approved a Monday report.
Four controls, and how to use them
Anthropic gave Enterprise admins four ways to hold this steady. Efficient mode keeps the merged experience on Sonnet and Haiku models. Model entitlements decide which roles can use which models. Spend limits work for the whole organization, a group or one person. Pooled group budgets give a group one shared monthly amount.
Efficient mode matters more than it looks. Sonnet 5.5 costs half as much per token as Opus 5.5, at $2 and $10 per million, so the same Monday report falls to about $5 and the 40 leads to about $10,400 a year. Token prices are also falling: Opus 5.5 came in 20% cheaper than Opus 5, so the same job costs less each time a new model arrives.
A pooled budget fits this kind of work better than personal caps. A team lead should not hit a wall in the middle of a report, but the group needs a ceiling and a way to see what used it.
Not everyone gets the merge yet. Organizations that use zero data retention for Claude Code, a HIPAA-ready configuration or customer-managed encryption keys stay on their current setup, so some regulated organizations will wait.
If your company is on Enterprise, ask before the rollout reaches your team which group you are in, what its monthly amount is and which models it can use. A manager who knows the amount can hand over big jobs without worry. One who learns it from the invoice usually ends up asking people to use Claude less.