For most of the last two decades, software had a simple price tag: a fixed fee per user, per month.
That model made software spend easy to reason about. Count the people, multiply by the price, and you knew the bill before it arrived.
AI is changing that. More of what businesses pay for is now priced by how much they use it, not by how many people have access. Spend management built around counting seats was never designed for that kind of bill.
Why seats were easy to manage
Seat-based pricing had three properties that made it manageable, even when it was wasteful.
- The cost was known in advance. The contract set the price, and the invoice matched it.
- Waste was easy to describe. An unused licence or a leaver who still has an account is a clear, fixable problem.
- Control was a yearly event. Most decisions happened at renewal: add seats, remove seats, change tier.
The whole discipline of software spend management grew up around these properties. Find idle licences, remove duplicate apps, get ahead of renewals.
How pricing is changing
AI costs real money every time it runs, so vendors are increasingly pricing by what it does rather than who can log in. A few models are now common.
- Pay per token. AI providers such as OpenAI and Anthropic charge for API use by the volume of text processed, at different rates for different models.
- Credits and allowances. Many tools bundle a monthly pool of AI usage into a plan, then charge for anything above it.
- Pay per outcome. Some vendors charge per task completed, such as a support query an AI agent resolves.
- Hybrid plans. A seat fee for access, plus usage charges on top for the AI features.
Seat-based pricing isn’t disappearing. But more and more software bills now have a usage component, and that part behaves very differently.
Why counting seats isn’t enough
Each of the properties that made seats manageable breaks down under usage pricing.
- The cost isn’t known in advance. It depends on what people and systems do during the month.
- Waste is harder to see. There’s no idle licence to point at. Overspend looks like an expensive model used for a simple job, or an automation nobody remembers setting up.
- Control can’t wait for renewal. By the time a yearly review comes round, a usage bill has already been paid twelve times.
A tool that only counts seats will show you who has access. It won’t show you what they used, which model they used, or what it cost. That’s exactly the information a usage bill depends on.
What finance teams and their advisers need now
Managing software spend now means managing two kinds of bill side by side. That calls for a few new habits.
- Track usage, not just access. For usage-priced tools, the useful numbers are consumption and cost by model, project and person.
- Watch it monthly, not yearly. Usage spend needs a regular check-in, with limits and alerts set ahead of the invoice.
- Give every cost an owner. Each project and key should belong to someone who can explain the spend.
- Keep one view of the whole stack. Seat-based tools and usage-based AI belong in the same picture, not in separate spreadsheets.
For accountants and bookkeepers, this is also an opportunity. Clients will increasingly ask why their AI line keeps moving. The adviser who can answer that question clearly is adding real value.
Software and AI spend, in one place
This is why AppVentory describes itself as Software & AI Spend Management. The two now sit on the same bills, and they need to be managed together.
AppVentory already helps businesses and their advisers find unused licences, duplicate apps and upcoming renewals. With AI Spend Management, you can also connect your OpenAI and Claude Developer Platform accounts, see spend by model, workspace, key and person, and set monthly limits and alerts before the invoice arrives.
Seats and usage, in one view.
See your full software and AI spend in AppVentory. Book a Demo with us today.
