Sometime in the next twelve months, one of your enterprise customers will open a renewal conversation with a sentence like this: "We have deployed agents on top of your platform. We need forty fewer licenses." The product did everything you promised. Adoption is healthy. Usage is up. And your revenue from that account is about to drop by a third.
Nothing went wrong. That is the uncomfortable part. Your customer did exactly what every board is telling every CIO to do right now - put AI agents to work. The agents log in, read, decide, and act. They just do not need seats. And your entire commercial model is built on seats.
Per-Seat Pricing Was Never About Seats
For twenty-five years, the seat was a proxy. One license meant one human, one human meant a unit of work, and a unit of work meant a unit of value. The proxy was never perfect - we all sold shelf-ware, and we all had power users who extracted ten times what they paid for. But it was close enough, and it had two properties CFOs on both sides loved: it was predictable, and it was easy to audit.
Agents break the proxy at its root. When an autonomous agent processes a cargo claim, reconciles an invoice, or triages a support ticket, work gets done without a human logged in. The volume of value flowing through your platform goes up while the number of humans touching it goes down. Seats and value have decoupled. Once that happens, per-seat pricing does not merely underprice your product. It actively rewards the customer for routing work away from licensed humans - which means it rewards them for shrinking your contract.
When one agent does the work of five licensed users, your pricing model is no longer measuring value. It is negotiating against you.
— Sethunath U N, Chief Consultant & Advisor, Mavis DxI have sat on both sides of this table across aviation, logistics, and healthcare deals. Buyers are not waiting for vendors to figure this out. Procurement teams are already modelling "agent-adjusted" license counts for FY27 renewals. If your pricing answer is not ready before that meeting, the answer will be written for you - as a discount.
The Three Candidates - and What Each Does to Your Revenue
There is no shortage of opinion on what replaces the seat. There is a shortage of honest accounting about what each option does to revenue predictability - the thing your investors, your board, and your own hiring plan quietly depend on. So let us take the three serious candidates one at a time.
You charge for the result: per claim settled, per shipment cleared, per reconciliation completed. It is the purest expression of value selling, and enterprise buyers find it genuinely attractive because your invoice now reads like their business case.
The catch is that outcomes are hard to define, harder to attribute, and hardest to audit. Who gets credit when the outcome involved your platform, the customer's data team, and a third-party agent? Every ambiguity becomes a dispute at invoice time. Outcome pricing works when the outcome is discrete, countable, and clearly yours - and turns into a quarterly argument when it is not.
Lowest. Revenue now moves with the customer's business volumes and with attribution disputes. Forecasting requires modelling their operations, not just your renewals.
You charge for what is used: API calls, workflow executions, agent actions, tokens, compute. It is honest - agents consume, so agents pay - and it scales revenue with the very automation that was eroding your seat count. The infrastructure players proved the model works at massive scale.
But consumption pricing transfers volatility onto both parties. Your customer's CFO loses budget certainty and starts demanding caps. You inherit a revenue line that swings with their seasonality - and a new obligation: you cannot price consumption you have not measured. Without a cost-to-serve model underneath, consumption pricing is guesswork with a decimal point.
Moderate, and it improves with instrumentation. Expect buyers to demand committed-use tiers with caps and collars. Your margin now depends on knowing your unit cost per action - precisely.
A fixed platform fee protects your revenue floor and funds the roadmap. Agents are packaged and priced as what they actually are - digital workers with a monthly rate, just like the human seats they stand beside. Metered usage sits on top for volume beyond the committed band.
This is where most mid-market ISVs will land, and for good reason: it keeps the CFO's budget line stable, it prices the agent instead of pretending it does not exist, and it lets you migrate account by account instead of betting the company on a single repricing event.
Highest. The platform fee anchors the forecast, digital-worker subscriptions grow as automation grows, and usage provides upside without carrying the whole P&L.
What Changes at the Negotiation Table
The shift is not only arithmetic. It changes the character of the renewal conversation itself.
- Customer's automation success shows up as your license shrinkage
- Discount pressure is the only lever procurement needs to pull
- You defend a headcount proxy you both know is broken
- Every agent deployed is revenue quietly walking out the door
- Customer's automation success grows your digital-worker line
- Conversation moves from price-per-seat to cost-per-outcome
- You present measured consumption, not estimated headcount
- Every agent deployed is expansion revenue with a named SKU
Do Not Switch Models. Earn the Right to Switch.
Here is where I will push against the panic. The worst response to this shift is a rushed, company-wide repricing announced in a quarterly release note. Pricing model changes fail not because the new model is wrong but because the ISV never built the foundation the new model stands on. Three things come first.
First, know your cost to serve. You cannot meter what you have not measured, and you cannot set a floor price on an agent action without knowing what that action costs you in compute, storage, and inference. If your cloud bill is still one undifferentiated line item, consumption pricing will amplify that blindness into negative-margin contracts. This is the same discipline I wrote about in the cost-to-serve piece - it has now moved from good practice to survival requirement.
Second, instrument before you invoice. Run the meters silently for two or three quarters. Learn the real consumption curves of your top twenty accounts before a single price is attached to them. When you finally present the hybrid model, you walk in with their own data - and nothing disarms procurement like precision.
Third, migrate deal by deal, not big bang. Start with new logos, where there is no anchor to defend. Move renewals opportunistically, leading with the accounts already deploying agents - they are the ones for whom the seat model is most obviously absurd, and they know it. Grandfather the rest until the story is proven.
The Question Underneath the Question
The seat was always a stand-in for something we found too hard to measure: the work itself. For twenty-five years the stand-in held, because only humans could do the work. That era is closing. Agents force every ISV to answer the question the seat allowed us to avoid - what, exactly, are we charging for?
The ISVs that answer it deliberately will find that agents are not a pricing crisis at all. They are the first expansion motion in a decade that does not require selling a single new seat. The ones that wait for the renewal meeting will discover the answer has already been decided for them.
So look at your own price book tonight and ask the only question that matters: are you pricing the seats, or the work?
Is your pricing model ready for the agent era?
Our ROI & TCO advisory builds the pricing handbook, cost-to-serve model, and Deal P&L discipline that let you walk into an agent-adjusted renewal with your own numbers - before procurement walks in with theirs. Thirty minutes will tell you how exposed your current price book is.
Outcome-based · No-commit · Monthly engagement
You continue only when you see real value. That is our promise.
Start the Conversation →