SaaS Pricing Strategies — searchable YouTube transcripts & AI Q&A

Shift AI Europe 2026: AI pricing when a few customers burn most of the AI cost

The speaker argues that AI pricing fails when a few customers consume most of the AI cost, creating a second hockey stick of cost that can erase profits. He recommends using fair usage mechanisms, such as setting a kill point and introducing deliberate service degradations (breaks) to control costs, and he reports that this approach improved one client's gross margin by 22 points in four weeks. He also notes that pricing should capture the spiky value some customers derive, as Figma did by introducing credits, but warns that symmetry between cost and value is not guaranteed.

Pricing Page unPacked - LOVABLE: Two $50 plans and credit top-ups

Lovable's pricing page deliberately uses a multi-pool credit system where both the Pro and Business plans cost $50, but the Business plan offers half the credits, making it effectively a more expensive tier. The speaker argues that this structure, along with credit rollovers and top-ups, drives conversion and usage, though it may create friction for larger customers who pay a premium for features that seem insufficiently differentiated. The enterprise add-ons, such as forward-deployed engineers, are seen as a way to increase consumption and adoption, especially for less technical customers.

How to deal with the double hockey stick problem in AI pricing

The speaker argues that AI pricing problems stem from mismatched value and cost hockey sticks, and the solution depends on which of four scenarios applies. For extreme cost spikes with flat value, fair usage mechanisms—like throttling or metering—can cut AI costs by 50-60% and boost gross margin by 15%, as shown in a case affecting under 600 users out of a million. When value is also spiky, you need both fair usage and sophisticated usage-based pricing to capture that value.

The "double hockey stick" problem in SaaS AI pricing

The double hockey stick problem is the main monetization challenge for AI products: value is concentrated in a few users, but so is cost, and the two are not necessarily aligned. The speaker illustrates this with a client where 0.1% of users drove 75% of AI costs, and the most unprofitable customer cost 400 times more than they paid, requiring an 800x price increase just to break even. He argues that flat pricing fails to capture the value hockey stick while leaving you exposed to the cost hockey stick, and that the relationship between cost and value is unstable over time.

Pricing Page unPacked - CLAY: actions, data credits and rollovers.

Clay's pricing strategy centers on separating data credits from actions, with actions becoming the primary monetization metric as data commoditizes. The speaker argues that by pricing actions, Clay captures value from workflow sophistication rather than just data volume, and suggests that offering data for free could be a bold competitive move. However, the current model introduces friction through separate pools and limited rollovers, which may need adjustment.