Optimisation

What the organisation is actually doing with the AI it pays for, read across every seat and every vendor that reports. Findings are things you could act on this week; the tables underneath are the evidence for them, and the bottom of the page names what we still cannot see.

Seats reporting
5
of 9 registered
Tokens / 7d
384.6M
across the estate
Tool runs
1.0k
14 distinct tools
True cost / 7d
$0.0000
$372.88 at list price, not billed

What we found

  • Decide

    2 different models across 3 seats

    Claude Code (don.p@naturebaby.com) → claude-opus-5; Claude Code (amy.w@naturebaby.com) → claude-sonnet-5; Claude Code (Claudia.Z@naturebaby.com) → claude-sonnet-5. Same product, different tiers. Worth deciding whether that split is intentional or just how each person happened to be set up.

  • Watch

    55% of activity comes from one seat

    Claude Code (don.p@naturebaby.com) accounts for 55% of tokens across the estate. Either the rest of the team has not adopted it yet, or this person is doing work the others could be.

  • Decide

    4 registered seats have reported nothing this week

    Cowork (don.p@naturebaby.com), Cowork (isobel.a@naturebaby.com), Cowork (Claudia.Z@naturebaby.com), Cowork (jacob@naturebaby.com). Registered, still on the books, no activity in the window.

  • Watch

    9 tools are used by more than one person

    Grep (172 runs across 2 seats), Bash (143 runs across 3 seats), Edit (56 runs across 2 seats), Read (29 runs across 2 seats), Write (28 runs across 2 seats), TaskUpdate (27 runs across 2 seats), TaskCreate (12 runs across 2 seats), ToolSearch (10 runs across 2 seats), Skill (5 runs across 2 seats). Repeated practice across people is where a shared skill or template pays for itself.

  • Decide

    2 tools average over 5 seconds

    Agent ~670.1s over 17 runs, WebSearch ~5.7s over 16 runs. Slow tools stall the whole turn, and they are usually the cheapest thing on this page to fix.

Model by seat

SeatModelCallsTokensShare
Claude Code (don.p@naturebaby.com)claude-opus-559085.4M22%
claude-sonnet-521035.0M9%
claude-haiku-4-5217.5k0%
Claude Code (amy.w@naturebaby.com)claude-sonnet-518034.4M9%
Claude Code (Claudia.Z@naturebaby.com)claude-sonnet-5187.4M2%
Claude Code (isobel.a@naturebaby.com)no calls in the window
Claude Code (james.o@naturebaby.com)no calls in the window
Cowork (don.p@naturebaby.com)no calls in the window
Cowork (isobel.a@naturebaby.com)no calls in the window
Cowork (Claudia.Z@naturebaby.com)no calls in the window
Cowork (jacob@naturebaby.com)no calls in the window

Tools across the estate

ToolRunsSeatsAvgRejected
WebFetch45713.4s
Grep1722102ms
Bash14331.8s
Edit56222ms
Read29272ms
Write28227ms
TaskUpdate27217ms
mcp_tool241812ms
Agent171670.1s
WebSearch1615.7s
TaskCreate12212ms
ToolSearch1023ms
Skill5211ms
SendUserFile42465ms

Rejected counts the times a person declined an edit the agent proposed. They are not failures – they are the only place a harness tells us a human pushed back, which is the control plane working.

Where a human stood in the way

Decisions
72
a person accepted or refused
Refused
1
of those 72
Awaiting approval
0
queued right now

1 of 72 decisions went against the agent.

What this cannot see yet

Discovery

Needs a connector into a system of record

Finding AI nobody registered means reading SSO logs, browser telemetry or a SaaS admin API. Telemetry only ever shows what already reports to us, which is the opposite problem.

Entitlements

Needs a scopes read per harness

Whether an agent can reach more than it needs is a comparison between what it was granted and what it has used. We hold the second half; the first has to be pulled from each vendor.

Benefit

Needs telling us what the work was worth

Hours saved and deals closed are not observable from a token counter. The cost side above is measured; the return has to come from you, per use case.