Find the AI seats nobody uses. Before the renewal does.
Companies buy AI seats in bulk and learn at renewal that a third were never opened. GhostAudit sets the licences you pay for against the usage its sensors already measure, and shows which seats are active, barely used or idle, with an estimate of what releasing them would save.
Example workspace. The Licences screen, with illustrative names, prices and figures.
- Active
- Low use
- Inactive
- Not monitored
| Licence | Seats | Use | Inactive | Est. savings |
|---|---|---|---|---|
ChatGPT Enterprise $60 per seat per month | 180 | 121 active · 22 low use | 31 | $1,860Per month |
Microsoft 365 Copilot $30 per seat per month | 250 | 128 active · 41 low use | 74 | $2,220Per month |
GitHub Copilot Business $19 per seat per month | 85 | 71 active · 6 low use | 8 | $152Per month |
Claude Team $25 per seat per month | 40 | 29 active · 5 low use | 4 | $100Per month |
Savings are estimates: sensors only see enrolled devices, and seats nobody can measure are left out of the figure.
- Measured over
- 30, 60 or 90 days
- Seat statuses
- Active, low use, inactive, not monitored
- New data collected
- None
- Plans
- Included in every plan
Seats you pay for, next to the use that was actually measured.
You record each paid AI licence once: the tool, the plan, the number of seats, the price per seat and the renewal date. GhostAudit joins that register to the usage its sensors report and gives each licence a row: how many seats are in active use, how many are barely used, and how many showed no use at all in the period.
Add the list of people who hold a seat, by hand or as a CSV export from the vendor's admin console, and the answer becomes specific. Each seat holder is one of four things:
- Active. Used the tool on enough days, for long enough.
- Low use. Opened it, but below the threshold you set. The default is under 30 minutes or fewer than 3 active days in 30 days.
- Inactive. No use seen in the period.
- Not monitored. None of this person's devices has a sensor that could see the tool. GhostAudit says so and does not guess.
The page also counts people who use the tool without being on the seat list. They are often on personal or free accounts, which is a governance finding of its own.
- Thresholds for low use are yours to set, per licence
- Renewal date and owner on each licence
- Seat lists export as CSV, with each person’s status in named mode, for whoever manages the vendor console
Example workspace in named identity mode. Names are fictional.
| Seat holder | Status | Active time | Last active |
|---|---|---|---|
Priya Raman Finance | Inactive | 0 min | Never seen |
Tomas Keller Sales | Inactive | 0 min | 12 Aug |
Dana Whitlock Operations | Low use | 14 min | 29 Oct |
Ines Carvalho Legal | Not monitored | n/a | No sensor on a device |
Marcus Oyelaran Engineering | Active | 11 h 20 min | Today |
A savings figure with its arithmetic shown.
The estimate is deliberately simple, so finance can check it: inactive seats multiplied by the price you entered. Low-use seats are listed for review and are not counted as savings, because the right response to someone who tried a tool twice may be training, not removal.
Where there is no seat list, the estimate is seats purchased minus the number of people seen using the tool, multiplied by the price.
It is an estimate, and the page says so. Sensors see enrolled devices only. Someone who uses their seat on an unmanaged laptop or a phone looks inactive. Seats held by people with no sensor are reported as not monitored and left out of the figure entirely. Treat the list of inactive seats as the people to ask, not as a list to cut without asking.
email,name [email protected],Priya Raman [email protected],Tomas Keller [email protected],Dana Whitlock
It works in pseudonymous mode, without ever producing a name.
Licence utilisation uses the usage data you already have. Nothing new is collected from any device, and nothing about what anyone typed is involved: the measure is time in the tool.
How specific the result is depends on the identity mode you chose for your sensors.
In named mode, each seat holder is matched to their usage by work email, and you see a status per person.
In pseudonymous mode, the emails on your seat list are hashed with your organization's salt, exactly as the sensors hash identities on the device, and matched to the stored hashes. Only totals per licence are shown: so many active, so many inactive. Which person is in which group is never produced, and for a seat list of fewer than five people no totals are shown at all, because in a group that small a total would identify someone.
In anonymous mode, seats purchased are compared with the number of devices seen using the tool.
| Identity mode | How seats are matched | What you see |
|---|---|---|
| Named | Each seat holder is matched to usage by work email. | Per person: active, low use, inactive or not monitored. |
| Pseudonymous | Seat-list emails are hashed with your salt, exactly as sensors hash identities, and matched to the stored hashes. | Totals per licence only. Never which person is in which group. Nothing for lists under 5 people. |
| Anonymous | Seats purchased are compared with the number of devices seen using the tool. | Seats in use and unused, as counts. |
What this can and cannot tell you.
We would rather you know the edges before you take a number to your CFO.
It measures use, not the account. Sensors report that ChatGPT was used on a device, by hostname or app name. They do not see which account was signed in. A person using a personal ChatGPT account on a work laptop counts as active against a ChatGPT seat.
Assistants built into other apps are only partly visible. Microsoft 365 Copilot is seen when someone uses the Copilot site or the Copilot app, not when they use Copilot inside Word, Excel, Outlook or Teams. For those seats, use Microsoft's own usage report alongside GhostAudit. GitHub Copilot is seen through the desktop agent, by its API connections, so it needs the agent and not only the browser extension.
The tool has to be in the catalog. A licence for a tool the sensors do not recognise can be recorded, and is marked as not measurable.
The seat register is yours to keep. GhostAudit does not connect to vendor admin consoles or billing systems. Seats, prices and seat holders are what you enter or import.
Our guide to finding unused ChatGPT, Copilot and Claude seats covers the vendors' own reports and how to combine them with sensor data.
Licence utilisation at a glance
- Register
- Tool, plan name, seats purchased, cost per seat per month, currency, renewal date, owner and notes. Fifteen currencies; totals are shown per currency, never converted.
- Seat holders
- Added by hand or by CSV import (up to 20,000 rows). Exportable as CSV, with each person's status in named mode.
- Periods
- 30, 60 or 90 days. Low-use thresholds are defined per 30 days and scale with the period.
- Statuses
- Active, low use, inactive, not monitored. A licence with no sensor that can see its tool is shown as not measured yet, with no estimate.
- Savings estimate
- Inactive seats multiplied by cost per seat. Without a seat list: seats purchased minus people (or devices) seen, multiplied by cost.
- Identity modes
- Per person in named mode. Totals only in pseudonymous mode, and none for lists under five people. Seats against devices in anonymous mode.
- Reporting
- Included in the AI usage report, with seat holders named only in named identity mode.
- Plans
- Included in every plan. See pricing.
Questions
- Does it connect to OpenAI, Microsoft or Anthropic to read our seats?
No. You record licences and import seat lists yourself, usually from a CSV export of the vendor's admin console. Usage comes from the sensors on your devices. We state this plainly because some buyers assume a direct integration.
- How is "inactive" decided?
A seat holder is inactive when no sensor reported any use of that tool by them in the selected period. If none of their devices has a sensor that could see the tool, they are shown as not monitored instead, and are not counted as savings.
- Can we use this without identifying individuals?
Yes. In pseudonymous mode you get totals per licence and never a per-person list. In anonymous mode you get seats against devices.
- Will it tell us what people use the tool for?
No. GhostAudit measures which tool and for how long. It does not collect prompts or content, so it cannot say whether a seat is used well, only whether it is used.
- Does it take a cut of the savings?
No. Licence utilisation is part of every plan at the normal per-employee price.
Check your seats before the next renewal.
Start a 14-day trial, enroll the people who hold seats, and record one licence. Usage starts to appear with the first sensor reports, and the picture fills in over the following weeks. No card is needed to start.