Notion now presents itself as a “Swiss of AI”, according to the words of David Tibbitts, Product Marketing at the publisher, whom BDM interviewed about the company’s strategy. Presented by Notion as its tenth employee, he has seen the tool evolve from a team wiki to a platform organized around AI agents, with the launch of its Agent in September 2025, then Custom Agents in February 2026.
From team wiki to the “context layer”
For David Tibbitts, AI tools share with Notion the blending of personal and professional use. He nevertheless sees a design difference: “Most AI tools we see today were fundamentally designed from the start as a solo, individual experience. Notion, on the other hand, was built from the outset as a collaborative platform. Any AI output you get in Notion lands on a shared surface. I think that’s a subtle but important difference.”
The executive acknowledges that the way the tool describes itself has varied a lot. Notion first presented itself as a connected workspace, sold per seat for its SaaS software, before AI’s arrival changed the narrative. “Overnight, from the sales perspective, we felt Notion was saying: forget all that, now it’s agents, agents, agents. But the message was more like: go present the agents to customers, help them build them,” he says. He is cautious about competition: “Everyone is fighting over the agent layer right now. It’s a very competitive space.”
To stand out, the publisher highlights its foundations, which it describes as a “backpack” for enterprise context. Collaboration features, granular permissions, version management or databases, he says, “benefit Notion’s agents, but also any agent you use in your company.” The idea is that a client can switch agents without rebuilding everything.
A “Swiss of AI” facing the labs
This positioning assumes working with labs that also offer their own agents. David Tibbitts embraces this ambivalence: “When Anthropic or OpenAI releases a new model, it’s often available in Notion within minutes. So yes, they are competitors in a sense, but they are also incredible partners.”
The publisher says it works with other providers besides those two, notably with some open‑weight models, and claims an agnostic approach. This diversity, he says, allows routing queries to the model best suited for the task. He justifies the choice by the multiplicity of offerings: “Do I send it to Claude Haiku 5.5 or to GPT‑6.1 Sol? Nobody wants to research every model. We do the homework for you.”
Pricing in credits under cost pressure
AI pricing has also changed several times. Notion AI became available as a paid option, billed per member, in February 2023. In May 2025, AI was included in the Business plan for new subscribers. In May 2026, Custom Agents switched to credit‑based billing, at $10 per 1,000 credits, and in October 2026, Workers, a building block of the developer platform, were likewise billed in credits.
David Tibbitts explains these shifts by the growing capabilities of agents. Twelve months ago, he says, the Notion agent could operate autonomously for a maximum of five minutes, versus up to 24 hours today. These advances have a downside: “Costs are exploding. And Notion doesn’t want to be a company that turns $2 into $1, as many of our AI peers seem to do.” Notion therefore continues to sell seats, as well as access to its AI services.
For usage‑based billing, the publisher emphasizes the visibility and control given to customers, who can decide, for example, which people can create agents or which costly models are allowed. The auto mode, which leaves model choice to Notion, should also allow paying according to the work done—a lightweight model sufficient for tweaking an article, a more powerful model for an ambitious goal. “Yes, it will consume more tokens, but we want to be sure we deliver the best possible experience, whatever your objective.”
