Are software seats disappearing? Eight pricing pages say no: usage is being built into them

Kristen Berman of Irrational Labs uses her product teardown newsletter to analyze the pricing pages of eight companies: Asana, Figma, Snowflake, Replit, Claude, ChatGPT, Wispr Flow and Granola. Her starting question is whether seats (per-user pricing) are disappearing in the AI era. Her conclusion is that she doesn't see them going away. What she observes is that usage is being bundled into seats, and that companies are putting a lot of effort into helping customers anticipate how much they will spend.

The underlying issue is a tension between two needs. The customer wants to know what next month's bill will be. The vendor needs to know whether that bill will cover what the customer consumes. AI complicates both. According to the author, the key is how much people hate surprise bills.

Berman argues that large consumer and B2B companies such as Asana, Notion and Figma are still selling seats, with features and usage allowances included. The mass migration to a fully usage-based model hasn't happened yet. Even so, these companies are innovating in how they build usage pricing into the seat. Asana counts requests rather than tokens, a far more understandable term for anyone trying to figure out what they are buying. Figma organizes its seats by role and offers several usage tiers within each, centering its value proposition on how much usage a person needs to do their job. Notion offers a free trial of its AI features on its lowest paid tier so users can try its Agent, and possibly as an incentive to upgrade.

Claude and ChatGPT also combine seats and usage, and neither offers a fully predictable bill. Claude Enterprise and ChatGPT Enterprise charge per seat for access and bill usage separately, so customers have to estimate their consumption. Claude Team and ChatGPT Business sell a standard $20 seat and a premium $100 seat. That gives two usage tiers to slot employees into and makes the bill somewhat easier to project, as with Figma. But it still requires guessing how many seats of each type are needed, and the author thinks people are likely to get it wrong at first. She hopes companies will be more generous with automatic downgrades if a customer miscalculated, and automatic upgrades when needed, rather than charging steep overage fees.

The companies best known for betting on a fully usage-based model are Snowflake, Clay and Replit. Berman notes that they exist but are not yet the majority. Being pioneers has cost them: all three have had roller-coaster years. Clay announced a very public price increase, Replit introduced "effort-based pricing" and Snowflake has invested heavily in sales teams capable of forecasting. According to the author, it makes sense that others aren't rushing toward pure usage pricing after seeing how hard it is to get right.

The second part of the piece stresses that pricing isn't just cost management. Willingness to pay is fundamental and, in her view, is being forgotten because the AI conversation revolves around cost. Before AI, it was clear what the paid plan offered: Canva puts locks on its paid templates, and Slack keeps 90 days of history and charges for more. Now almost anything can be replicated with AI, so those features are far less defensible, and most companies are searching for the feature that will move someone from free to paid.

Wispr Flow and Granola offer two answers. In Wispr Flow, the paid plan includes an admin panel, centralized billing and a shared workspace. In Granola, centralized billing and MCP access, with stronger security on the enterprise plan. Berman admits that none of these features is revolutionary, but they are requirements for businesses. Her rule: people don't want to pay you, so a paid feature has to be truly exceptional or mandatory.

The author adds that all of this will change. When she started giving talks on pricing, the story was add-ons: Notion led with a $6 AI add-on, and now AI is included because it's built into everything. That cycle, she says, is the new normal. Pricing becomes an ongoing conversation between product, marketing and whoever owns the number, and companies that treat it that way will have an easier time than those still waiting for the annual review.

Note on the material: the text received is the newsletter article, not the video in which the author walks through the eight pricing pages. This summary therefore covers only what is explained in writing. The article is an opinion piece based on observing public pricing pages, not a quantitative study, and provides no adoption or revenue data.