Per-seat pricing: why agentic AI kills the seat

September 16, 2026 · 5 min · Free, with full transcript

Per-seat SaaS pricing assumes a human logs in and does the work. Agentic AI does not log in, so one user can run forty background agents that never touch the interface. This episode argues the seat is already dead as a billing unit, and that charging agents a seat fails in both directions: too cheap and you give away a thousand humans of work, too expensive and the customer builds it in-house. You come away with a way to pick a new unit. The discussion covers outcome pricing, Intercom's roughly 99 cents per resolution, Salesforce Agentforce at about $2 per conversation, efficiency deflation as inference costs fall, and why credible attribution decides who keeps pricing power.

Listen to the podcast

4 min

Chapters

Key takeaways

  • Stop counting seats and count value events per account: tickets resolved, reports shipped, money moved.
  • If your value-event count is rising while seat count stays flat, reprice against that event before a competitor does.
  • Meter a business result the customer would have paid a human far more for, not tokens, tasks or invoices scanned.
  • Treat vendor figures like Intercom's 99 cents per resolution, Salesforce's $2 per conversation and OpenAI's tenfold annual inference cost drop as marketing, and cross-check them.
  • Build the ability to prove your agent caused the outcome, because whoever measures it credibly owns the pricing conversation.
Read the full transcript

Host:Welcome to the Leaders Insights Podcast. Today's episode, how Agenic AI breaks the math behind your SaaS pricing model. Every SaaS company on earth charges per seat. You're about to tell me that's broken. Convince me.

Expert:It's not broken. It's dead. It just hasn't stopped moving yet. Per seat pricing assumes a human logs in and does work. Agenic AI, software that plans and executes multi-step tasks on its own, doesn't log in. One user now spins up 40 background agents that never touch your interface. You're billing for chairs in a room nobody sits in anymore.

Host:Come on, vendors could just charge the agents a seat too. Problem solved.

Expert:They tried. It falls apart instantly. A human sales rep might send 50 emails a day. An agent sends 50,000. If you charge that agent one seat, you're giving away a thousand humans worth of work for the price of one. If you charge it like a thousand humans, the customer laughs and builds it in-house. The unit e-inter nido, the thing you actually meter, stopped being a person. And most pricing pages haven't noticed.

Host:So what's the real unit now?

Expert:Work done. Outcomes. The number of resolved tickets. Closed leads. Reconciled invoices. Intercom saw this early. They moved their support product to charge per resolution, roughly 99 cents, when their AI actually closes a case. Now, that's Intercom's own framing. And they sell support software. So take the 99 cents as marketing, not gospel. But the direction is right. You bill for the result, not the login.

Host:Here's the uncomfortable part. If I charge per outcome and my AI gets better, I close each outcome cheaper. So I make less money as my product improves. That's insane.

Expert:That's the trap everyone walks into. And it's the smartest thing you asked. Yes, efficiency deflation is real. As the model gets cheaper, your per outcome margin can rise, but the price the market tolerates falls faster. OpenAI's own numbers show inference cost, what it costs to run the model once, dropping something like tenfold per year. That's their figure. They sell the tokens, so cross-check it. But even discounted, the trend means the thing you're metering keeps getting cheaper to produce and cheaper to buy. Volume has to do the heavy lifting.

Host:So the answer is just sell more units?

Expert:That's what every dying business says. No. The answer is to meter value the customer feels, not work you perform. There's a difference. O'Reilly's research on AI adoption makes the point well. The companies keeping pricing power aren't charging for tokens or tasks. They're charging for a business result the customer would have paid a human far more to get. If your agent recovers $100,000 in unpaid invoices, nobody argues about a $4,000 fee. If you charge them per invoice scanned, they'll haggle you to zero.

Host:But that requires the vendor to prove they cause the outcome. Attribution is a nightmare.

Expert:It is, and that's the moat now. Whoever measures the outcome credibly owns the pricing conversation. Salesforce is betting the whole company on this with Agentforce, charging around $2 per conversation their agents handle. Their number, they obviously want agents everywhere, so treat it skeptically. But notice what they did. They picked a countable event a customer can see happening. The vendors who survive aren't the ones with the best model. They're the ones who can stand behind a bill and say, here's what we did, line by line.

Host:And the ones who can't?

Expert:They keep selling seats to a workforce that no longer exists. It's like charging admission to a stadium after everyone started watching from home, technically a business briefly.

Host:All right, someone's running a SaaS company today per seat, watching agents eat their model. What do they do Monday morning?

Expert:Pull your usage data and find the one moment in your product where a customer clearly gets value. A ticket resolved. A report shipped. Money moved. Count how often it happens per account, not per user. If that number is climbing while your seat count is flat, your pricing is already lying to you. Reprice against that event before a competitor does it for you. The unit was never the person. It was the outcome. It always was. We just had humans in the way hiding it.

Host:Sources for today's episode, the decoder, TechCrunch AI, OpenAI, vendor, AI Lab, Kated Nuggets, O'Reilly Radar. That's your five. The full AI track, structured, no fluff, is at MBA-training.com.

Read and go deeper