AI Pricing Models After Per-Seat: A SaaStr Podcast Review
Key takeaway
This episode of The Agents is worth an hour for anyone who sells or prices SaaS. Its best material is a live disagreement between Jason Lemkin and pricing consultant Ulrik Lehrskov-Schmidt on whether to charge for AI from day one or launch it free first, plus a concrete look at how credit pricing can triple a small team's bill. The pricing numbers are the hosts' own estimates, the guest's firm sponsors the episode, and the second half drifts into SaaStr's own agent stories.
Podcast: The Agents, episode 18, from SaaStr Hosts: Jason Lemkin, founder of SaaStr, and Amelia Lerutte, SaaStr’s Chief AI Officer Guest: Ulrik Lehrskov-Schmidt, founder of the pricing consultancy Willingness to Pay Length: about 68 minutes, on the SaaStr AI YouTube channel
AI pricing models are where software sales gets hard. Buyers are used to paying per seat. Vendors are adding agents that cost real money to run, and they want to charge for usage, credits or outcomes. Sales teams sit in the middle, explaining bills that can change shape from one renewal to the next.
This episode of The Agents, SaaStr’s show about running a company with AI agents, takes that problem on directly. The first half is a pricing conversation with Ulrik Lehrskov-Schmidt, whose firm has run more than 200 B2B pricing projects, according to the episode. The second half returns to the hosts’ own agent stack. This review was written from the full episode transcript, and it separates the operating lessons from the opinions and the sponsor message.
What the episode says about AI pricing models
The episode opens with a confession. Lemkin and Lerutte both cancelled Canva subscriptions, independently, and both say they had not logged in for months. Lemkin calls it “stealth churn”: usage drops to zero long before the cancellation, and a vendor that is not watching never sees it coming. Their bigger point is that their AI agents had already picked other tools for images and slides and had no reason to switch back. If an agent does not pick you early, the hosts argue, it may never reconsider.
From there the episode has four parts:
- HubSpot and credit pricing. Lerutte prices out a small marketing team under HubSpot’s old and new models.
- Pricing strategy with Lehrskov-Schmidt. Where to raise prices, Figma’s move to credits, and whether to charge for AI at launch.
- SaaStr’s AI VP of marketing. How the hosts’ agent, called 10K, helped close a sponsorship deal and lift renewals.
- Agents and security. Agents calling each other, an agent ordering flowers without asking, and why the hosts will not build their own CRM.
The HubSpot math
The most concrete segment is Lerutte’s comparison. She models a three-person marketing team with 10,000 contacts, a weekly newsletter of about 40,000 emails a month and 20,000 workflow actions a month. Under the old model, she says, the bill was about $1,300 a month, mostly seats. Under the new model, seat costs drop from about $800 to about $300, but workflow actions become credits beyond an included allowance, and she puts the total at about $3,700 a month.
These are the hosts’ own estimates, not figures from HubSpot, and they do not walk through every line item. But the shape matches what Lehrskov-Schmidt says happened to his own company as a HubSpot customer: his head of marketing called to move the newsletter elsewhere, and they negotiated a discount. His critique is not that HubSpot moved to credits. It is that one credit charges the same for a newsletter send and an AI-driven outbound action. He suggests two credit pools instead, one for data enrichment and one for actions, an approach he associates with Clay.
For a sales team, the lesson is practical. When a vendor’s pricing model changes, the customers who get hurt most are the ones whose usage pattern the vendor did not model. Those are the accounts that call to churn.
The strongest points
Raise prices where value is already high, not across the board. Lehrskov-Schmidt argues that a flat 10% to 30% increase is the way to hurt a customer base. The better move, he says, is to find accounts where the product already delivers far more value than they pay for and repackage so those accounts pay more, while others pay the same or less. He puts it memorably: almost every company has a $50,000 account that would pay a million.
The “double unlock.” His best example is Figma. As he tells it, Figma launched AI features, drove adoption, met them with credits and only then monetized. The new consumption model captured the cost of AI and also exposed value hidden in the seat model, where some customers would pay far more than they did. He thinks many seat-priced products are sitting on the same hidden value.
The disagreement on sequencing. This is the most useful part of the episode because the two do not agree. Lemkin calls Figma’s AI building product uncompetitive and says teams should be forced to charge for AI from day one, at prices as high as the category leaders, so the market gives a clear verdict. Lehrskov-Schmidt pushes back with a client case: a company of about $30 million in revenue that charges for a new agent immediately and gets low adoption cannot tell whether the product or the price failed. Launching free first, he says, separates the two questions. Both arguments hold up, and which one fits depends on how sure a team is that its agent is good.
An agent that doubles contract value. Lemkin describes a portfolio company at about $30 million in annual recurring revenue that launched an agent with no free version, priced to roughly double contract value, and saw about 3% of customers adopt it in the first week. His challenge to founders is to build agents worth 50% to 100% of the current contract, not small AI features.
Speed beats process. Lehrskov-Schmidt says he used to recommend pricing committees and now recommends a single “pricing dictator”, because similar clients take nine months or four weeks to roll out a credit system depending on how many layers must agree. Lemkin adds a warning about a common panic move: forcing every monthly customer onto annual contracts, which he says looks good in a spreadsheet but mostly delays churn and adds collections work.
Where it is thin or a pitch
The guest’s firm sponsors the episode. The ad read for Willingness to Pay’s AI pricing ebook runs at the start and end, and the guest is its founder. His advice is reasonable and specific, but listeners should know the segment doubles as promotion.
The numbers are anecdotes. The HubSpot figures are the hosts’ model. The Canva and Figma margin claims come from Lemkin, stated from memory. The doubled contract value and 3% adoption come from an unnamed company. None are verified on the show.
The second half is about SaaStr. The story of 10K co-selling a sponsorship with Base44, the Wix-owned app builder, is a vivid example of an agent preparing decks, pricing custom work and filling in vendor paperwork. Lerutte says renewal rates for six-figure sponsors rose from the 20s to the 70s in percentage terms, and even she says the agent called its own contribution “operational support”. It is an interesting case, but it is one company’s account of its own tools.
The security segment speculates. Lemkin discusses an open source penetration testing tool being turned against companies and guesses at the cause of one breach. His advice, not to store sensitive customer data in a homemade CRM, is sound. The specifics are his speculation.
What a sales team can do with it
- Map each customer’s usage against the new model before renewal. If your company is moving to credits, find the accounts whose bills jump most and talk to them first. HubSpot’s story shows the call to churn comes fast.
- Sell the agent on outcomes, not actions. Lemkin’s example is a vendor whose agent books four meetings in its first week. That is a value story a buyer can take to finance. “Per action” pricing without that story invites pushback.
- Watch for stealth churn. Ask customer success for accounts with falling logins well before renewal, and treat AI agents as users too. If a customer’s agents stop calling your API, that is a warning sign.
- Give one person the pricing call. If deals stall on pricing exceptions, push for a single owner who can decide fast.
- Price your own AI tools with the same eyes. The same logic applies when your team buys AI sales tools or an AI SDR: know which meter you are on before the bill changes.
Who should listen
Founders, revenue leaders and anyone who owns pricing or packaging will get the most from the first 35 minutes. Account executives and customer success managers dealing with renewals under a new pricing model will find the HubSpot segment useful for empathy and talk tracks. People building agents into sales work, such as a GTM engineer, may prefer the second half, which pairs well with the earlier Vercel lead qualification talk review. Teams weighing which AI agents for business to buy will hear a useful buyer view of agent pricing throughout.
Verdict
This is one of the more useful public conversations on AI pricing models because it shows two experienced people disagreeing in specifics. Take the numbers as illustrations, skip the sponsor reads, and keep the core lessons: charge for agents that deliver obvious value, find hidden value before raising prices across the board, and move faster than the competitors in your category.
FAQ
What is the episode about?
It is episode 18 of The Agents from SaaStr, where Jason Lemkin and Amelia Lerutte discuss AI pricing models with Ulrik Lehrskov-Schmidt of Willingness to Pay, then talk about their own AI agents, renewals and security.
Is per-seat pricing dead?
The episode argues it is fading for products that run AI agents, because agents cost money to run and do work that seats do not measure. Lemkin still says buyers prefer seat pricing for many products that were always seat based.
What is the double unlock in AI pricing?
Lehrskov-Schmidt uses the phrase for a move to consumption pricing that both covers the cost of AI and captures value that was hidden in the old seat model, using Figma’s move to credits as his example.
Should a SaaS company charge for AI features at launch?
The episode does not settle it. Lemkin says charge from day one to get a clear market signal. Lehrskov-Schmidt says launching free first can separate product problems from pricing problems.
Who should listen to this podcast episode?
Founders, revenue leaders and pricing owners should listen to the first half. Sales and customer success teams handling renewals under new pricing will also find it useful.
Sources
Checked November 19, 2026.