Adopt
Stack

What your AI stack costs, and what it returns

Seat counts come from each vendor's admin console. Whether the work stuck comes from Adopt. Neither number decides anything on its own — a seat can be open every day and still be the wrong tool for every task that person has.

Demo workspace — synthetic organisation. Not measured telemetry.
Reclaimable, annualised
€1,286,208
5,054 seats paid for and inactive
Annual stack cost
€3,182,160
8 tools
Seats licensed
12,800
across 8,800 people
Seats active
7,746
61% of what you pay for
Seats earning their keep · 7,746Idle · 5,054

Seat activity is seeded, not measured. These figures demonstrate the method. Connect a vendor API on Connections and this page recomputes from your own admin data.

Per toolRanked by money sitting idle
Microsoft 365 Copilot
Microsoft
4,200
63%
61%(94)
€528,192
Keep

63% seat activity, 61% task adoption. Holding steady.

Gemini for Workspace
Google
1,100
8%
40%(10)
€267,432
Cut

Only 8% of 1100 seats show any activity. 1,013 seats are paid for and untouched.

ChatGPT Enterprise
OpenAI
2,000
57%
50%(20)
€260,100
Keep

57% seat activity, 50% task adoption. Holding steady.

Slack AI
Salesforce
3,400
73%
63%(8)
€97,956
Keep

73% seat activity, 63% task adoption. Holding steady.

Notion AI
Notion
600
43%
—
€41,400
Keep

43% seat activity. Holding steady.

GitHub Copilot
GitHub
420
66%
86%(7)
€32,376
Scale

66% of seats active and 86% task adoption. This one earns its money.

Claude Enterprise
Anthropic
900
90%
89%(44)
€30,912
Scale

90% of seats active and 89% task adoption. This one earns its money.

Cursor
Anysphere
180
36%
83%(6)
€27,840
Keep

36% seat activity, 83% task adoption. Holding steady.

Overlapping coverage

Tools strong in the same categories. Redundancy in an AI stack hides well, because the products have different names and the same job.

Microsoft 365 Copilot and Gemini for Workspace€1,701,600

Both strong across 5 categories. The figure is what you currently spend on the pair, not a saving — consolidating would recover part of it, and the Ledger says which of the two the work actually sticks to.

ChatGPT Enterprise and Claude Enterprise€902,400

Both strong across 4 categories. The figure is what you currently spend on the pair, not a saving — consolidating would recover part of it, and the Ledger says which of the two the work actually sticks to.

ChatGPT Enterprise and Gemini for Workspace€890,400

Both strong across 2 categories. The figure is what you currently spend on the pair, not a saving — consolidating would recover part of it, and the Ledger says which of the two the work actually sticks to.

Gemini for Workspace and Claude Enterprise€592,800

Both strong across 2 categories. The figure is what you currently spend on the pair, not a saving — consolidating would recover part of it, and the Ledger says which of the two the work actually sticks to.

Uncovered work

Categories no licensed tool serves well. These are the places work leaves the stack.

Every task category has a capable licensed tool. Where work still leaves the stack, the cause is routing or policy rather than a missing purchase — see Shadow.