Notion Focuses on a "Context Layer" and AI Credits

Key Takeaways

1Notion's AI features are largely transitioning to a credit-based pricing model, with $10 for 1,000 credits on Custom Agents.

2Notion claims an agnostic approach to models and automatic routing of requests to the deemed appropriate model.

3The publisher highlights a collaborative "context layer," beyond just the competition of agents.

Why it matters : Notion is changing both its product positioning and its business model to integrate AI in a collaborative, multi-model, and usage-based billing manner.

Notion outlines a strategic shift: the publisher claims a role as a "context layer" that is agnostic between agents and team data, while generalizing a credit-based billing system for its AI functions. David Tibbitts, Product Marketing, describes rapid integrations with models from labs and automatic query routing to the model deemed appropriate.

Extended AI Credits: $10 for 1,000 Credits on Custom Agents

Notion has made several pricing adjustments around its AI functions. After introducing Notion AI as a paid option billed per member in February 2023, the company included AI in its Business offering for new subscribers in May 2025. In May 2026, Custom Agents transitioned to a credit-based billing system, priced at $10 for 1,000 credits, and in October 2026, the platform's Workers also switched to this model. David Tibbitts describes AI pricing as a rapidly evolving and sometimes chaotic area, noting that the entire industry is still searching for the right billing formula.

Asserted Agnosticism and Rapid Availability of New Models

Notion states that it works with OpenAI and Anthropic, while also offering some open-weight models, in an approach presented as completely agnostic. This strategy includes routing user AI queries to the model deemed suitable for each task. David Tibbitts claims that partnerships allow newly released models from labs to be available in Notion within minutes, with some users reporting they have seen them arrive in Notion before ChatGPT or Claude. The labs are thus described as both competitors and partners.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

A Bet on Task Routing Rather Than a Single Model

David Tibbitts believes that few users want to compare the diversity of available models themselves, citing examples like Claude Haiku 5.5 and GPT-6.1 Sol. He argues that limiting oneself to a single provider misses out on capabilities available elsewhere. Notion therefore claims to "do the homework" for teams, studying the strengths and weaknesses of models to determine the right tool for each task.

Collaboration by Default: Results and Agents Shared with the Team

The company contrasts its collaborative approach with AI experiences designed primarily for individual use. In Notion, results generated by AI appear on a shared surface, and agents can be adjusted with teammates. David Tibbitts places this approach in a context where many AI tools are used for both personal and professional tasks.

Notion Claims an Evolution of the Workspace Towards Agent Integration

Notion recalls its historical foundation as a connected workspace, sold on a SaaS basis for team wikis, project management, and documentation. Following the arrival of AI, the directive has been to present agents to customers and help them build their own, while recognizing that the "agent layer" is a highly contested space. The publisher now emphasizes a "context layer" built on ten years of development, such as collaboration, granular permissions, version management, and databases. David Tibbitts describes it as a "backpack" for context, skills, and memory, designed to avoid having to rebuild everything when switching agents and to remain collaborative by nature. This repositioning coexists with the launch of the Agent in September 2025, Custom Agents in February 2026, and the adoption of credit-based billing for certain functions.