Superhuman · AI product · Mid · 45 to 60 minutes

How should Docs AI move beyond chat?

Superhuman is a productivity software company best known for a fast, premium email app. Answer this product management case and get it graded line by line against a weighted rubric.

product strategy
AI UX
prioritization
technical judgment
risk management

The situation

Your GM for Docs asks you for a recommendation on the next phase of AI in Superhuman Docs. The first version shipped quickly and proved there is demand for AI help inside documents, but the current setup behaves mostly like a chat layer. It can answer some questions, yet it struggles when the task requires seeing structured content like tables, understanding the document's building blocks, or making precise edits directly in the doc. Users and internal teams now want an assistant that feels native to editing, not bolted on. Engineering says they can keep extending the current system, but each new capability is taking longer and feels more brittle. A deeper platform change could unlock better actions and reasoning, but it would absorb time that could otherwise go to user facing features and carries delivery risk. Leadership needs a recommendation now because planning for the next product cycle is underway, and they must decide whether to keep improving the current AI path or invest in a new foundation for Docs AI.

What to produce

Write a 1 to 2 page product strategy memo.

Your answer must

State the core product problem and why it has become urgent now.
Define the key user jobs for Docs AI beyond chat, including at least one example involving structured content and one involving direct document edits.
Compare at least two credible paths, such as extending the current approach, a hybrid path, or a deeper foundation change, with tradeoffs across user value, speed, quality, and future flexibility.
Recommend one path and outline a phased execution plan for the next product cycle, including what you would do first.
Propose success metrics and leading indicators that would show whether the chosen path is working.
Name the biggest risks, including user trust or quality risks, and how you would de-risk them before broad rollout.

Constraints

Base your reasoning only on the context given, do not assume proprietary data, external partnerships, or a breakthrough model.
Your recommendation must preserve Superhuman's bar for speed and user trust, the AI cannot feel slow or make silent, hard to review changes.
Keep the scope to one product area decision for Docs AI, not a company-wide reorganization or brand new product.
Do not assume the team can pause all Docs feature work for an open-ended period.

Format

Use five headings: problem, user jobs, options, recommendation and plan, metrics and risks.

How Superhuman actually played it

A Superhuman senior PM explains why the team restarted Docs AI from scratch, beginning with the MCP, so the AI can read tables off screen, edit content directly and understand building blocks. The post links the rebuild to lessons from the earlier AI launch. A useful case on when to rebuild an AI foundation.

Rebuild the foundation when early AI hit its limits
Design the tool interface for AI, not for people
Define what the AI should do beyond chat before building

Built from The Making of Superhuman Docs: How We Rebuilt the AI in Superhuman Docs From the Ground Up by Bharat Batra.