LEONThe seat is empty
Paying per seat was always a stand-in for value. AI just broke the link between the two.
You pay for software per person, per month. That is how software has always been priced and always will be.
The seat was always a stand-in
Vendors needed a number that grew with the customer, and headcount was the easiest one to find. That works while people are the ones doing the work. When one person with an AI agent does the work of several, the count of logins says nothing about the value delivered.
The irony: a vendor ships an AI feature that makes each user more productive, and the per-seat model bills fewer seats.
Agents don’t buy seats
Gartner has predicted that about 40% of enterprise apps will include task-specific AI agents by the end of 2026, up from under 5% in 2025. Agents don’t log in or show up in the admin panel, but they use the product heavily. That is why vendors are shifting to usage-based, credit-based and outcome-based pricing.
Seat bill vs work done
What replaces the seat
You pay per task, message or API call. It’s fair when usage is predictable and scary when it isn’t.
You prepay a pool of credits. It’s flexible, but the conversion rates are easy to hide.
You pay per resolved ticket or booked job. This lines up best with value, but it’s the hardest to measure.
One price for the whole shop. It’s simple and predictable, and it doesn’t punish you for growing.
At your next renewal
The vendor asks you to commit to the same seat count for 3 years. Smart move?
Rewritten from the original No Sacred Cows post for the LeonLink blog.
New builds, monthly. One email when something new drops. No spam, unsubscribe anytime.