Notion · AI product · Mid · 45 to 60 minutes

How should Notion launch Custom Agents?

Notion is a workspace app where people create docs, wikis, projects, and workflows for individuals and teams. Answer this product management case and get it graded line by line against a weighted rubric.

product strategy
prioritization
pricing
ai product judgment
metrics

The situation

Akshay and the leadership team have asked you, the PM for Custom Agents, to recommend how Notion should move the product beyond beta. Early beta feedback is promising: some teams are using agents for internal Q&A, release note updates, support reply drafts, and recurring research across Notion and connected tools. But usage is uneven. Some workspaces are only testing one simple agent, while others are starting to rely on agents for important operational workflows. The launch decision is hard for three reasons. First, model costs vary a lot by workflow and model choice, so Finance wants a monetization plan that matches value without slowing adoption. Second, workspace admins want visibility and guardrails before they support broader rollout, especially if agents run in the background. Third, the product team believes the builder should stay usable for non engineers, not only AI specialists. You need to propose a launch plan that balances growth, trust, and cost discipline, and makes clear what Notion should ship now versus later. Assume you have 8 weeks to prepare the beta exit, with the current team and no major billing platform rebuild.

What to produce

Write a 1 to 2 page product strategy memo.

Your answer must

Recommend the target user and the 2 to 3 workflow types Notion should optimize for first, with rationale based on the beta context.
Propose a pricing approach, including the charging unit, why it fits customer value and Notion's cost structure, and one meaningful alternative you considered.
Define the minimum admin controls and visibility features required at launch, and explain why each is necessary.
List the 3 to 5 metrics you would use to judge launch success across adoption, quality, and economics.
Name the biggest risks, including at least one adoption risk and one trust or spend risk, and give a mitigation for each.
State one thing you would explicitly not launch yet, and why.

Constraints

The experience must remain usable for non engineers who know the workflow best.
The plan must work for team and company level use, not just for an individual power user.
Your recommendation should account for variable inference costs across different models and tasks.
You cannot rely on adding a large new operations layer for admins to manage agents day to day.

Format

Use sections titled recommendation, target use cases, pricing, launch scope, metrics, and risks.

How Notion actually played it

After a Custom Agents beta, users created over a million agents in two months and teams went from first agent to automating end-to-end workflows. Notion also announces usage-based pricing and says agents got up to 50 percent cheaper to run. A concise look at what a beta taught about adoption and pricing.

Beta usage shows which workflows matter.
Lower cost per run enables pricing changes.
Fast teams treat agents as internal products.

Built from What we learned during the Custom Agents beta by Akshay Kothari.