Claude Enterprise Cowork Pilot Plan
Decide what happens on 16 November, pilot unified chat and Cowork by role, keep spend and retention under control, and bring outcomes to your steering group.
What's inside this template
Who it's for
CIOs, AI Operations Leads, security and finance teams running Claude Enterprise at organisations where Chat and Cowork unified is not yet switched on
When to use
Before Chat and Cowork unified turns on by default on 16 November 2026; when planning a pilot by role; before expanding the unified experience to more teams
Key benefit
A deliberate decision instead of a default, with roles, settings, spend limits and retention gaps handled before the switch, and evidence to expand
Sections included
- Pilot details
- The decision
- Role map
- Settings review
- Retention gaps
- Spend guardrails
- Pilot measures
- Opportunity log with ICE scores
- Outcomes
- Expand, hold or stop
- Message to staff
- Steering group lines
- Implementation notes
Complete template content
To use this plan, copy it with the "Copy page" button above or open the Google Doc version. Read the companion analysis, what the 16 November Cowork switch-on means for your Claude Enterprise rollout, for the thinking behind it.
Claude Enterprise Cowork pilot plan
A free template for CIOs, AI Ops leads, security and finance teams who run Claude Enterprise. Use it to decide what happens when Chat and Cowork unified turns on by default on 16 November 2026, run a pilot by role, and bring evidence to your steering group before you expand. It works on its own; you don't need any platform to use it.
0. Pilot details
- ☐ Organisation:
- ☐ Decision owner (usually the AI Ops lead):
- ☐ Sign-off (usually the CIO):
- ☐ Decision date (before 16 November 2026):
- ☐ Chat capability on? (yes / no)
- ☐ Cloud code execution and file creation on? (yes / no)
- ☐ Cowork or Cowork in the cloud on today? (yes / no)
- ☐ Zero data retention, HIPAA-ready or customer-managed encryption keys? (If yes, the unified experience isn't available yet. Record that and stop here.)
- ☐ Sandbox organisation available? (yes / no)
1. The decision
Pick one and write down why.
| Option | Choose when | Our choice (tick one) | Why |
|---|---|---|---|
| Enable now | Cowork in the cloud is already reviewed and in wide use | ||
| Pilot by role (recommended for most) | You want evidence before everyone gets it | ||
| Keep disabled | Security review not started, or a known blocker |
If you keep it disabled, set a date to revisit:
2. Role map
Before you turn the organisation setting on, move anyone who shouldn't have it yet to a custom role with Chat and Cowork unified turned off. New custom roles start with all capabilities, so check them too.
| Role or group | Members (count) | Chat and Cowork unified (on / off) | Efficient mode (on / off) | Models allowed | Spend limit | Wave |
|---|---|---|---|---|---|---|
| Example (fictional): Finance pilot | 25 | On | On | Sonnet, Haiku | Pooled monthly budget | 1 |
| Everyone else | Off until wave 2 | 2+ |
3. Settings review
Agree each setting before the pilot starts. Check the names against Anthropic's current admin guide.
| Setting | Our choice | Why | Owner |
|---|---|---|---|
| Network egress (off, package managers only, custom domains, all) | |||
| Connector policies (blocked, needs approval, per-role) | |||
| "Always allow" for connector tools | |||
| "Skip all approvals" mode | |||
| "Automatically approve" mode | |||
| Share chats and share cloud sessions | |||
| Remote control of members' computers | |||
| Require Trusted Devices | |||
| Scheduled tasks | |||
| Custom data retention period | |||
| Compliance API and audit review | |||
| Desktop device policy (MDM) |
4. Retention gaps
Custom retention covers conversations and the cloud sessions Claude starts from them. Decide how you'll handle what it doesn't cover.
- ☐ Cowork cloud sessions a member starts directly (kept until the member deletes them). Our approach:
- ☐ Scheduled tasks and their runs (keep running after the conversation that created them is deleted). Our approach:
- ☐ Cowork history stored on members' computers. Our approach:
- ☐ Memory of members who leave. Our approach:
- ☐ Members told that existing memory moves to a new memory system, and the old memory may be deleted 14 days later.
5. Spend guardrails
- ☐ Pooled group budget for the pilot (monthly amount):
- ☐ Member-level spend limit:
- ☐ Efficient mode on for the pilot role? (yes / no)
- ☐ Stop rule. If pilot spend passes this amount in a week, we will:
- ☐ Reports updated to include the Chat and Cowork unified line in every spend, token and message total
- ☐ Headcount taken from overall active users, not by adding products together
6. Pilot measures (two to four weeks)
| Week | Active pilot users | Unified spend | Tasks returned as finished work | Issues or incidents | Approval prompts (too many / about right / too few) | Notes |
|---|---|---|---|---|---|---|
| 1 | ||||||
| 2 | ||||||
| 3 | ||||||
| 4 |
7. Opportunity log with ICE scores
Each task people hand over and get back as finished work is a candidate opportunity. Score impact, confidence and ease from 1 to 10 and multiply them for a score out of 1,000.
| Opportunity | Raised by | Process it changes | Expected outcome | Impact (1-10) | Confidence (1-10) | Ease (1-10) | ICE score |
|---|---|---|---|---|---|---|---|
| Example (fictional): monthly board pack first draft | Finance pilot | Month-end reporting | Two days to half a day | 8 | 6 | 7 | 336 |
8. Outcomes
Baseline each process before the team relies on the unified experience, then measure again.
| Process | Baseline (before) | After (week 4) | Basis (estimated / sampled / measured) | Who confirmed it |
|---|---|---|---|---|
9. Expand, hold or stop
Expand to the next wave only when all of these are true:
- ☐ No unresolved security or data incident from the pilot
- ☐ Spend stayed within the pooled budget, or the overspend is explained by value
- ☐ Retention gaps have an agreed approach
- ☐ At least one outcome is sampled or measured, not just estimated
- ☐ Pilot users would object if it were switched off
Decision and date:
10. Message to staff (draft)
From [date], Claude can do more than answer questions. Ask it for the finished thing: a report, an analysis or a deck. When a task needs it, Claude carries on in a secure cloud workspace and keeps going if you close your laptop. We're starting with [pilot group] and expanding in waves. Please don't put [restricted data classes] into Claude. Tell [AI Ops lead] what you hand over and what comes back, so we can back the best uses.
11. Steering group lines
- Decision on the 16 November default: [enable now / pilot by role / keep disabled], because [reason].
- Pilot: [group], [members], [dates], within a budget of [amount].
- So far: [number] opportunities logged, top ICE score [score] for [opportunity].
- Ask: approval to expand to [wave 2 group] on [date], or [decision needed].
12. Implementation notes
- Check every setting against Anthropic's admin guide before you rely on it. Names and defaults change often.
- Run the pilot in a sandbox organisation if you have one.
- Usage data can take up to two days to appear in analytics.
- If you log opportunities and outcomes in Kowalah, the free plan covers that.
- This is a planning tool, not legal or data protection advice. Agree data rules with your DPO and legal team.
Get started
How to use this plan
Check your starting point
Confirm whether Chat and Cloud code execution and file creation are on, and whether you use ZDR, HIPAA-ready or CMEK. That tells you if 16 November affects you.
Make the decision
Choose enable now, pilot by role or keep disabled, write down why, and get it signed off before 16 November.
Map roles and set guardrails
Move non-pilot staff to a custom role with the capability off, then set Efficient mode, a pooled budget and member limits for the pilot.
Review settings and retention
Agree egress, connector, approval, sharing, remote control and scheduled task settings, and decide how to handle the retention gaps.
Run the pilot and decide
Track measures weekly for two to four weeks, log and ICE-score opportunities, baseline outcomes, then expand, hold or stop.
Questions
Frequently asked questions
Common questions about the unified Claude experience on Enterprise
What happens on 16 November 2026?
Can we stop it switching on?
Who gets it when it's on?
How do we control cost?
Does our data retention setting cover everything?
Is this legal or data protection advice?
More templates
Related templates.
Ready when you are
Want help before 16 November?
Kowalah is a UK Anthropic Implementation Specialist. We help CIOs and AI Ops leads roll out Claude Enterprise features with the controls on: roles, spend limits, retention and a pilot that produces outcomes. If you'd like help, book a conversation.