"Everything will be consumption-based."
You hear this a lot in SaaS right now. It’s the current wisdom. The idea is that in the age of AI, charging for a "seat" is archaic. Why pay for a human login when the AI is doing the work?
It sounds logical. It sounds fair. But when you look at what the actual market leaders are doing, the story changes.
If you peel back the layers of the fastest-growing AI tools—Cursor, Lovable, Rovo, Copilot—you don’t see a pure consumption model. You see a convergence. Everyone is landing on the exact same hybrid spot.
Why? Because economics are stubborn.
The Problem with "Pure" Anything
Pure consumption pricing feels fair until you try to get it approved.
Finance teams hate surprises. They want to know if the bill in November will look like the bill in October. A model where a rogue engineer can accidentally spend next quarter’s budget over a weekend isn't "flexible"—it’s a liability.
On the flip side, pure seat pricing feels stable until your users actually start using the product.
In the old days of SaaS, a "power user" just clicked more buttons. It cost you basically nothing. Today, a power user on an AI tool can torch more compute in a single afternoon than an entire team used to consume in a month. If you stick to a flat seat price while your users lean heavily on GPT-4 or Claude 3.5 Sonnet, your margins don't just shrink; they evaporate.
The Evidence
The market has already solved this. The solution isn’t to kill the seat. It’s to cap it.
Look at the pricing structures of the heavy hitters:
Cursor: You pay $20/month for a Pro seat. That sounds like a standard subscription. But look closer. That seat comes with 500 "fast" premium requests. Go over that limit, and you’re throttled to slow requests or pushed to pay-as-you-go. It’s a seat, with a consumption ceiling.
Lovable: They charge $25/month for the Pro plan. But that’s not really for "access." That $25 buys you exactly 200 monthly credits (and nobody knows what that means). If you want to build more, you wait for next month, buy more credits or you upgrade. It’s usage disguised as a membership.
Atlassian Rovo: The new kid on the block charges $20/user/month. But again, read the fine print. That seat includes 2,000 "Rovo credits." If your team goes wild with agents, you start paying $0.01 per extra credit. The seat gets you in the door; the credits keep the lights on.
GitHub Copilot: For Pro, it's $10/user/month. But even Microsoft, with all its compute power, has limits. The Pro tier gives you 300 "premium requests" per user. Want more? You need the Pro+ tier at $39/user to get 1,000 requests. The tier isn't about features; it's about volume.
Predictable Cost, Protected Margins
This hybrid model—Seat + Cap—is the compromise that lets everyone sleep at night.
For the customer, it provides the predictability finance teams crave. They can model a $20/headcount cost. It fits in the spreadsheet.
For the vendor, it creates a safety valve. You can offer an "unlimited" feel for 90% of your users, but you have a contractual way to stop the top 10% from eating your infrastructure alive.
It’s not exciting. It’s not a "paradigm shift." It’s just practical.
Our View from the Hill
At Released, we are still on a pure seat-based model.
For now, it works. The compute intensity isn't at the level of generating thousands of lines of code (yet).
But we aren't blind. As we layer more AI into categorizing and analyzing customer feedback, the gravity of compute costs will eventually pull on us, too.
There is a temptation to over-engineer pricing early. To build complex metering systems before you even know how people use the tool. We try to avoid that kind of uphill pushing.
Better to start simple. Let the ball roll downhill. Right now, a simple seat price removes friction and lets teams get started. But we know that eventually, fairness will mean change. Not because we want to squeeze every penny, but because sustainable tools need sustainable economics.
Seat pricing isn't dead. It just got a new best mate.
