What's the Bet: Notion
Has the time for AI agents finally come?
Author’s note: This is the latest in my “What’s the Bet” series. AI has the potential to change everything, and founders have to make company defining choices in response. That is not easy. In this series, companies pay me to analyze their bet and explain it to you. I retain editorial independence. Today’s bet is brought to you by Notion, a software product that is central to The Leverage functioning, so it is one I care about quite a bit. If you have any feedback, just reply to this email.
Notion’s bet: an Agent that finishes cross‑app work.
Most AI products today are trapped inside a single surface—a doc bot, a slide bot, an email bot, or even a browser bot. This is useful for automating an individual task, but brittle the moment the task spills across apps, draws in multiple people, or requires structural changes. Notion’s wager is one Agent that plans and executes end‑to‑end workflows across your tools and your workspace. While individual specialization can almost certainly have more functionality, Notion thinks that people would rather stitch together workflows versus perfecting them.
What this Agent does that single‑app copilots can’t:
Plans & executes multi‑step work (Search → Read → Decide → Update → Notify), not just generates a reply.
Moves across surfaces: turn meeting notes into tasks, update a roadmap database, notify owners in Slack/Teams, draft the stakeholder email all with one request and one result.
Builds structure within Notion on the fly: The AI agent can create pages, design databases, and reorganize information to make the workflow reusable.
Learn your playbook: profile pages + memories capture filing rules, tone, references, and shortcuts so the Agent behaves more like a teammate.
Bridges platforms: pulls from Notion, Google Drive, Slack, Gmail, MS Teams/SharePoint/OneDrive, and the web, then pushes updates back. This allows it to close the loop on tasks or communication happening outside of Notion.
The technical advantage
Under the hood, Notion is a block‑based workspace: a document is a tree of addressable blocks (paragraphs, headings, lists, embeds). Databases are collections of pages with typed properties (selects, dates, formulas, relations/rollups). This matters because the Agent can query page properties, update databases, and refactor child‑blocks across dozens of pages in one pass. Having a centralized technical language gives the agent a higher success rate then if it has to switch between file types and applications.
Why this architecture unlocks agentic work
1) Workflow > document.
Most AI copilots produce content inside a file. Notion’s Agent manipulates structures (databases, relations, filtered views) and content, so it can “compile” a workflow: e.g., define owners, create filtered dashboards, set due dates, and send the comms—then run that workflow repeatedly.
2) Cross‑tool closure.
The Agent sits atop your “work graph” and closes loops across tools: it doesn’t just read Slack and Drive; it reconciles that context back into your Notion system of record and notifies stakeholders.
3) Compounding leverage curve.
The bet is that while today Agents will result in some quick time saves, that effort will compound. On day 1 that might mean time saved on generating one report. By week 4, the agent will have a reliable, reusable flow done by creating a custom Agent and a database schema for it to reference. By the second quarter, there will be multiple, custom Agents running on schedules, keeping your roadmap, docs, and comms current. The payoff compounds because the Agent invests in infrastructure you keep.
4) Governance built‑in.
Because the Agent works where the permissions live, it naturally inherits access controls, leaves an auditable trail (plan + citations), and gives you an undo safety net—far easier than policing a zoo of Zapier automations glued across file silos.
Competitive landscape
Incumbents: Google Workspace & Microsoft 365.
Gemini/Copilot are useful inside Docs/Word, Sheets/Excel, Slides/PowerPoint, Gmail/Outlook. But they’re fundamentally file‑first. Cross‑app workflows often need scripts, connectors, and manual glue. Notion’s Agent advantage is workspace‑first execution: one content graph the Agent can reason over and change in one pass.
All‑in‑one challengers: Coda, Airtable, ClickUp, Monday, etc.
There are many new age productivity companies that are doing some version of blending docs + data with strong AI features and automations. Notion’s edge is its block model, deep database ergonomics , and MCP bridges that let the Agent operate across your stack.
For the first category of incumbents, Notion will have to compete with some of the most scaled GTM organizations in the world (very hard) and for the second, with startups that are just as nimble and furiously innovative (also very hard). Still, Notion is the first one to fully integrate agents across the productivity suite like this and as such, have first mover advantage.
A brief history of Notion (1.0 → 3.0)
Notion 1.0 — Pages (August 2016): The original “all‑in‑one workspace” released on Product Hunt with a page‑ and block‑centric editor.
Notion 2.0 — Databases (March, 2018): Databases became first‑class with multiple views (table/board/calendar/gallery) and relations/rollups—turning Notion from flexible notes into a system builder.
Notion 3.0 — Agents (September 18, 2025): Agents become AI teammates that plan and execute across your workspace and connected tools—announced around the Make with Notion 2025 event.
How to use it and how I tried it out
Open Agent from the corner of any page.
State the goal (“Turn these notes + tickets into a weekly status with owners and dates,” “Aggregate customer feedback and update the roadmap + notify PMs”).
Review the plan & citations, approve, and let it run.
Iterate: teach it your filing rules and tone via its profile page. You can then spin up Custom Agents for recurring work.
For me, I have a big database of podcast guests, their stages of outreach, etc. Using the agent, I was able to easily transform that database into a content management system, saving me hours of low-level thinking. I’m now thinking about how I can automate its access to everything, soon my work will be steering multiple of these agents versus performing manual labor myself.
The simple promise
Most AI helps with work. Notion’s Agent finishes it—across apps, across structure, inside one workspace. If teams automate more of their jobs than ever before, Notion becomes the practical replacement for file‑first suites. The bet will truly start to pay off as agents become easily shareable within organizations, and more and more of the mind numbing admin work within a firm is automated.
If you are interested in trying out Notion Agent, you can try it out here! Even if you think this bet is wrong, I’d encourage you to try. I was surprised how much easy utility was available to me.
I run a 70 person consulting shop on Notion and we are extremely excited about the possibilities to be unlocked by Custom Agents. Some of the use cases were looking to test out are:
- Personal AI assistants for each team member to reduce the mental overhead of project management inside Notion
- Weekly reports sharing project progress with clients, especially around the recurring work our team does. Maybe even a Slackbot clients can interact with to ask their own questions of our Notion data?
- Updating statuses, archiving outdated items, and maintaining data accuracy across our library of client hubs & company wiki resources. Agents will keep our data up-to-date & make it more trustworthy/useful to team members