Fixed fee · Two weeks · Spend, adoption, overlap
Find out what the AI tools are costing and what they bought
When a renewal is due and nobody can say what the Copilot seats, the ChatGPT workspace and the coding assistants actually bought, our engineers answer it in two weeks for a fixed fee.
We pull seat-level usage from each vendor's admin console and API, set it against the contracts, find the tools doing the same job, and name the uses that changed how your teams ship. You get an estate map and a consolidation plan before the renewal.
- Length
- Two weeks
- Who it is for
- Finance and platform leaders holding a renewal they cannot justify
- Bench
- 115+ engineers, averaging 15 years each
Three years of buying has left most companies with Copilot seats, a ChatGPT tenancy, a coding assistant per team, and four pilots nobody measured. Finance holds the contracts, IT admins hold the Microsoft 365 usage reports, and the GitHub and OpenAI data sits behind APIs a platform team has to query, so nobody has the whole picture or the week it takes to put it together.
What we look at
- Every AI tool with a contract, a seat count, and a renewal date
- Seats bought against seats that logged in this month
- Where two or more products cover the same job
- Inference and API spend outside the seat licences, which is usually where the surprise is
- What sensitive data each tool can currently reach
- The handful of uses that genuinely changed how a team works, so they survive the consolidation
What you get
- A written estate map: tool, cost, seats, usage, renewal date
- The overlaps, with a recommendation on which to keep and why
- Renewal exposure for the next twelve months
- A consolidation plan, and where a built system would cost less than the licences it replaces
Where the usage data actually lives
Microsoft, GitHub and OpenAI each count an "active" seat differently, and two of the three can show a seat as idle when it isn't, so their exports only mean something side by side.
Microsoft 365 Copilot reports sit in the admin center under Reports, then Usage, then Microsoft Copilot. The per-user table gives a last activity date for each app, over 7, 28, 90 or 180 days, and exports to CSV. User names are hidden by default until an admin changes the reports setting, and only an intentional action counts, so opening the Copilot pane in Word without sending a prompt doesn't make someone active.
GitHub Copilot exposes every seat through the REST API, with last_activity_at, the editor it came from, and any pending cancellation date. GitHub's own documentation adds the caveat that matters at renewal: IDE activity only reaches last_activity_at if the user has telemetry enabled in their editor. A null there can be a privacy setting rather than an unused seat.
ChatGPT Enterprise and Edu workspaces export analytics as CSV from Workspace settings, up to twelve months at a time, refreshed somewhere between one and twenty-four hours behind. The Compliance API is the route to raw logs, and OpenAI says plainly that its counts won't match the analytics dashboard, because one is raw system data and the other is cleaned.
gh api --paginate /orgs/YOUR-ORG/copilot/billing/seats \
--jq '.seats[] | [.assignee.login, (.last_activity_at // "never"), (.last_activity_editor // "-")] | @tsv' \
| sort -k2Checked on 11 October 2026 against Microsoft Learn: Microsoft Copilot usage report, GitHub: REST API for Copilot user management, OpenAI: Compliance API vs user analytics and OpenAI: user analytics for ChatGPT Enterprise and Edu.
Seven steps from exports to a plan
01
Start the exports early
Admin-level usage exports take longest to get, because they need somebody with the right admin role in each tenant. We send the list of what's needed before the engagement starts, with the contracts or a list of them.
02
Build the estate map
Every AI tool with a contract, its seat count, its renewal date and what it costs, in one table. Inference and API spend outside the seat licences goes on the same table, since it's usually where the surprise is.
03
Set usage against seats
Seats bought against seats that did something this month, using each vendor's own definition of active and correcting for the ones that undercount, like a telemetry setting that hides real use.
04
Find the overlaps
Where two or more products cover the same job, which one each team actually reaches for, and what moving them to one would break.
05
Talk to the teams
An hour each with two or three of the teams using the tools. The uses that changed how a team works are found here, and they're the ones the consolidation has to protect.
06
Check what each tool can reach
Which repositories, drives and mailboxes each product can read today, because a tool nobody uses still holds whatever access it was given.
07
Write the plan
The estate map, the overlaps with a recommendation on which to keep and why, renewal exposure for the next twelve months, and where a built system would cost less than the licences it replaces.
Renewal homework
Questions you can answer from your own admin consoles
- Is every AI contract, with its renewal date, in one list?
- How many seats did something in the last 28 days, by each vendor's own definition?
- Before calling a seat idle, have you checked whether telemetry or report concealment is hiding its use?
- Which teams pay for API keys on a card that procurement never sees?
- Which two tools do the same job, and which one does each team open?
- What can each tool read today, and does anybody still need it to?
- Which use changed how a team works, and would it survive a cut?
FAQ
AI Tools Assessment: common questions
What is an AI tools assessment?
A fixed-fee two-week audit of the AI products a company already pays for: seats against actual usage, overlapping tools, inference spend outside the licences, and what each one can reach. It ends in an estate map and a consolidation plan. The page is /ai/tools-assessment.
Are you going to tell us to cancel everything?
No. Some of these tools are the best money in the budget and the point is to find which, so the ones that work survive the next cost review. Sometimes the right answer is more seats for the one tool a team actually adopted.
How do we review AI coding tool usage before a renewal?
Pull each vendor's seat-level export and put them in one table with the contract value and renewal date. For GitHub Copilot that's the seats endpoint of the REST API, which gives last_activity_at per seat; for Microsoft 365 Copilot it's the usage report in the admin center; for ChatGPT Enterprise it's the workspace analytics CSV.
Check telemetry and report concealment settings before calling any seat idle, because both can hide real use. Then ask the two or three heaviest teams which tool they'd keep if they could keep one.
Do you resell any of the tools you would be assessing?
No. We do not take a margin on any of them, which is the only reason an assessment like this is worth reading.
What do you need from us?
The contracts or a list of them, admin-level usage exports, and an hour each with two or three of the teams using the tools. The usage data is the part that takes longest to get, so it is worth starting before the engagement does.
The rest of our AI work
AI Readiness Assessment
Leaders with AI on the roadmap and no clear view of what is in the way
Agentic AI development
Engineering leaders whose agent pilot cannot get past review
AI assistants and enterprise search
IT and data leaders asked for a company assistant that can't leak
AI workflow automation
Operations and engineering leaders with a process that half-completes
AI enablement
Engineering leaders whose team needs to own this, not outsource it
AI transformation
Executives with a mandate, a budget, and fifteen pilots