Cost per AI Seat: How to Calculate the Real Cost

Cost per AI seat is the amount spent on each license of an AI tool assigned to a person. It is the most common measure and the most misleading: it equals the list price divided by seats, but ignores how many seats are used and for which tasks. The real cost comes from recalculating it per active user and per use case, including ancillary costs.

Why list price is not the real cost

List price answers the vendor's question: how much does a license cost. The CFO's question is different: how much does it cost to get value from a person who uses the tool.

At least four gaps separate the two figures:

• Seats purchased and not used. Every license assigned to someone who does not use it raises the cost of those that are used.
• Ancillary costs. Training, license administration, integrations and internal support do not appear in the fee, but they are part of the cost. The article on AI total cost of ownership covers them.
• Contractual commitments. Minimum volumes and term can keep seats on the books that nobody uses.
• Uneven use. Two active users do not produce the same value: the average cost hides who uses the tool every day and who opens it once a month.

For this reason the unit price is a starting point for negotiation, not a measure of efficiency. Whoever decides a renewal looking only at list price compares offers that do not describe their own usage.

Purchased seats versus active seats: the first correction

The first correction is to separate three numbers that are often confused:

1. Purchased seats: the licenses paid for under contract.
2. Assigned seats: the licenses linked to a person.
3. Active seats: the licenses used regularly in a defined time window, for example the last 60 days.

Between the first two there may be a share of licenses paid for and never assigned. Between the last two, a share of licenses assigned and never used. Both are cost without value. The definition of threshold and window is covered in how to find unused seats.

The formula is simple:

• Cost per seat = total annual spend ÷ purchased seats
• Cost per active user = total annual spend ÷ active seats
• Activation rate = active seats ÷ purchased seats

Cost per active user equals cost per seat divided by the activation rate. With an activation rate of 80%, the real cost is 25% higher than the nominal one.

Worked example: from cost per seat to cost per active user

The example is hypothetical: the values are invented for illustration and do not represent real prices from any vendor.

A company buys 500 seats of a tool at €300 per seat per year.

• Annual subscription spend: 500 × €300 = €150,000.
• Active seats in the last 60 days: 400. Activation rate: 400 ÷ 500 = 80%.
• Cost per seat: €150,000 ÷ 500 = €300.
• Cost per active user: €150,000 ÷ 400 = €375.

The gap is €75 per user, or 25% more (375 ÷ 300 = 1.25). The 100 inactive seats weigh €30,000 a year (100 × €300), which is 20% of spend.

Now add ancillary costs, again hypothetical: €10,000 a year for training, administration and support. Full spend rises to €160,000 and the full cost per active user becomes €160,000 ÷ 400 = €400, a third more (400 ÷ 300 ≈ 1.33) than the starting unit price.

From cost per user to cost per use case

Cost per active user says whether seats are used, not whether they are used well. To know whether the spend is justified you need cost per use case: how much each task the tool is used for costs.

The method has three steps. Assign each active user a main use case, multiply the number of users by the cost per active user, and compare the cost with the expected or measured value. Use cases come from a use case register or, failing that, from interviews with department heads.

Continuing the hypothetical example, with a subscription cost per active user of €375, the 400 active users are distributed as follows:

• Drafting text: 200 users, 200 × €375 = €75,000.
• Document summarization: 120 users, 120 × €375 = €45,000.
• Coding assistance: 80 users, 80 × €375 = €30,000.

The sum is €75,000 + €45,000 + €30,000 = €150,000, which is the subscription spend. Cost per use case does not come from a different price but from an allocation choice: the share of spend each task absorbs. The next step is to compare it with a benefit measured against a baseline, as the guide to use case ROI explains.

A use case that absorbs 20% of spend and produces no measurable result is the first candidate for review. One that absorbs a lot and generates clear value should be protected.

What data do you need and where to find it?

The calculation requires five data points. None is hard to obtain, but they are scattered across different systems.

• Spend per tool: contracts, orders and invoices, including purchases outside central procurement (company cards, department budgets).
• Purchased and assigned seats: vendor admin consoles and any IT spreadsheets.
• Usage per user: vendor admin consoles expose activity data to varying degrees. Check for each tool what granularity is available and how long it is retained.
• Role and department: the HR directory, to aggregate the analysis without looking at individuals.
• Ancillary costs: training budget, administration hours, integration costs, to be estimated with the owners.

To respect confidentiality principles, read the analysis by role and department. If usage is collected at individual level, Article 4 of the Italian Workers' Statute (Law no. 300 of 20 May 1970) on remote monitoring applies, to be checked with HR and the DPO before collecting data. Data aggregated by department is, in any case, almost always enough for the calculation.

How to use the real cost in decisions

Cost per active user and per use case serves three decisions: how many seats to keep at the next renewal, which tools to compare, where to invest in adoption before cutting. A department with a high cost per active user and many inactive seats has an adoption problem, not a price problem: the answer may be targeted training instead of a reduction.

A practical sequence:

1. Calculate cost per seat and cost per active user for each tool.
2. Rank the tools by the gap between the two values.
3. For those at the top of the list, check the cost per use case.
4. Decide case by case whether to activate, reduce or replace.
5. Repeat the calculation before every renewal.

For the full picture of reduction levers, see how to reduce AI license spend.

Where Delphis fits

Delphis is a vendor-independent platform that connects spend, adoption, impact and regulatory obligations. It reconciles costs with actual usage, calculating cost per seat, per active user and per use case across multiple tools, with data aggregated by role or department. It also lets you simulate the savings from a reduction and measure impact against a baseline. The analysis setup is described on the Spend page.