Market sizing with the assumptions kept

Research and writingFounder or analyst

We sized the market three times in a month. Each time an agent re-found the same developer counts and rebuilt the same sheet.

How do I stop redoing the same market sizing research every time I ask an AI agent?

Have the first agent save the numbers, sources and assumptions with ArcRoom, and later sessions reuse them through find_task. They only change the assumption being questioned, and decisions such as 'we count only teams of 5+' stay with the numbers. Numbers from the web need a date and an expiry rule.

Also asked as: My agent rebuilds the same market size sheet every time. How do I stop that? · How do I save market sizing assumptions so an AI agent can reuse them? · Can an AI agent change one assumption without redoing the whole market sizing?

The work

The first agent saves the numbers, the sources and the assumptions.

ArcRoom keeps
  • Reused answer
  • Decision
  • Saved decision
The next agent gets

find_task → reuse: "Teams that run AI agents at work: …, source …, assumption: 30% pay for tools"

Step by step

  1. The first agent saves the numbers, the sources and the assumptions.
  2. Later sessions reuse them and only change the assumption being questioned.
  3. Decisions such as 'we count only teams of 5+' stay with the numbers.

What the agent sees

find_task → reuse: "Teams that run AI agents at work: …, source …, assumption: 30% pay for tools"

Compared with today's tools

Sizing spreadsheet with sources and dates

Keep the sizing sheet with sources and dates, and attach it to the agent next time.

Setup
0 min
Ongoing
update the sheet and attach it each time

ArcRoom

Every agent reads the team playbook before it works and writes to it when it finishes. Nothing to update by hand.

Setup
2 min (npx arcroom init)
Ongoing
None

The sheet is the right record and may match ArcRoom; ArcRoom saves pointing the agent at it, and keeps the reasoning behind each number.

Where it stops

Numbers from the web need a date and an expiry rule.

Try it in your project

Works with the agents your team already uses, and any agent over MCP or the HTTP API.

More in research and writing