Notion Rebuilds AI Pricing Around Credits, Not Seats

Notion just ripped out its AI pricing model and replaced it with a credit system that supports bring-your-own-model, while Karri Saarinen is publicly wrestling with the same problem from the builder's chair: when AI makes it trivial to add features, what stops your product from turning into an incoherent pile of additive junk. ### Notion - Rebuilds Notion AI Around Credits and Model Routing, Not Flat Seats **Source:** https://www.notion.com/blog/notion-ai-more-for-less **Credibility:** High (first-party product announcement from co-founder Akshay Kothari) **What happened:** Notion rebuilt its AI product around a new credit-based system, smarter model routing, long-running agents, and the ability to bring your own AI (BYO model/API key) instead of being locked into Notion's bundled models. Co-founder Akshay Kothari frames it as "more for less" — the same or better AI capability at lower effective cost per task, driven by routing requests to the cheapest model that can handle them well rather than a single default model for everyone. **Key technical details:** - New credit system replaces (or supplements) the prior flat-inclusion model for AI usage - Smarter model routing: requests get sent to different underlying models depending on task complexity, rather than one model for all AI calls - Long-running agents: tasks can now execute over extended time horizons instead of single-shot completions - Bring-your-own-AI: users/orgs can plug in their own model or API key instead of relying solely on Notion-provided models - Ships alongside a related post on how Notion connected its AI to 160 new third-party apps in three weeks using an internal Notion agent to handle the Figma-style setup and testing work **Why it matters for PMs:** This is the tension between "AI feature as included perk" and "AI feature as metered cost center" playing out in real time. Flat-rate AI inclusion was a land-grab tactic — it got AI into every workspace without forcing a purchasing conversation. A credit system means Notion now has granular cost attribution and can tie usage to willingness-to-pay, but it also means power users will start watching a meter, which changes behavior. The BYO-model option is the more interesting signal: it's an admission that no single model wins on cost/quality for every task, and that locking users into one model is a liability as the model market gets more competitive and more commoditized. If you're pricing an AI feature, this is the shape of where things are heading — usage-based credits plus the option to swap in your own model, instead of a single AI tier at a fixed price. **Critical questions:** - Does moving to credits reduce AI usage among price-sensitive users, undermining the habit-formation Notion spent two years building? - How does "bring your own model" affect Notion's unit economics if their heaviest users opt out of Notion-provided models entirely? - Model routing improves cost efficiency, but does it introduce inconsistency in output quality across similar tasks, and will users notice? - What happens to the simplicity of Notion's pricing page once "it depends on which model and how many credits" becomes the answer? **Action you could take today:** If your product bundles AI into a flat subscription, pull usage data on your top 10% of AI users and model what a credit-based cost structure would do to their effective price. That's the group who will notice first.