Seat license

Anyone who doesn’t buy a program outright but instead is allowed to use it in exchange for an ongoing fee needs permission from the manufacturer to do so. This permission is called a license. With a seat license, this permission is counted per person: for every employee who is meant to use the program, the company pays a fixed amount per month or per year. The English technical term for this is “seat.” Twenty employees with access mean twenty seats and therefore twenty times the price. Whether someone uses the program eight hours a day or once a month makes no difference to the bill.

Why companies are billed by seats

For buyers, this model is above all one thing: predictable. A company with 300 employees knows exactly at the start of the year what the software will cost. This can be written into a budget and justified to management. Models in which the price fluctuates with usage are harder to plan for and, in the worst case, lead to nasty surprises on the invoice.

For the provider, the seat license has a different advantage. Revenue grows automatically along with the customer. If the company hires 50 new people, 50 new seats are added without anyone having to renegotiate. This is precisely why software companies talk so often about the number of seats per customer in their quarterly figures. This number shows whether a product is spreading throughout a company or remains stuck in a small department.

A common mistake is to confuse a seat license with a device license. A device license is tied to the computer: whoever logs in on that machine is allowed to use the program. A seat license is tied to the person and travels with them across laptop, phone, and office computer.

What a seat means technically

Behind every seat is a personal user account. The provider maintains a list of which accounts are paid for and checks against it at every login. Once all seats are occupied, the next employee gets no access until a seat becomes free or the company purchases more. There are often additional tiers as well: a basic seat can cost less than a pricier seat with extended features.

With AI tools, this creates a tension. The provider’s costs, after all, are not tied to people but to computational work. Every request to a language model costs electricity and computing time on expensive graphics chips. Experts call this operation of the finished model “inference.” A very active user can cost the provider more than their seat brings in.

That’s why many providers now mix the models. The seat includes an allowance, such as a certain number of requests per month. Anyone who needs more pays extra based on consumption. Such allowances are often calculated in credits or tokens, i.e., small billing units for processed text.

Seat licenses in products and financial news

This model is found wherever business software is sold. Microsoft 365, Slack, Salesforce, and Adobe bill per person. AI assistants such as Microsoft Copilot or the business versions of ChatGPT also follow this pattern, typically in the range of 20 to 30 euros per person per month.

The term appears in business news when analysts argue about the future of these companies. The concern is: if AI agents take over tasks that humans used to perform, the number of employees with access declines. Under this model, fewer people automatically means less revenue. Some providers are therefore already switching to prices based on tasks completed rather than on headcount.

For private individuals, the topic is less often relevant, since simple individual subscriptions usually apply there. But anyone procuring software for a club or a school quickly runs into the question: how many seats do we really need? Unused seats are the most common reason for unnecessarily high software costs.