The AI Research Engine — a repeatable path from question to recommendation

The problem

You research a market-entry question for two weeks. You have a folder of material and no clearer view. Nobody defined what would constitute an answer, so every source raised another question, and you stopped only because the meeting arrived.

What you need

Deciding: whether to open a second warehouse in Hyderabad. At stake: ₹40 lakh setup, ₹6 lakh/month running.

The tool

Any assistant

The prompt

What comes back

STAGE 1 — FRAMED Will a Hyderabad warehouse, costing ₹40 lakh to set up and ₹6 lakh a month to run, generate enough additional south-India contribution within 18 months to justify itself — primarily through faster delivery improving conversion and reducing returns? STAGE 2 — LOAD-BEARING CONDITIONS…

Verify before you use it

• Do condition (a) from your own data first. It is a pivot table, it takes an hour, and it can end the research.
• The ₹6 lakh monthly estimate is an assumption until tested. Note that the decomposition flags it rather than accepting it.
• Condition (d) — management overhead — is the one businesses systematically underestimate. Two conversations with peers is cheap and informative.
• Set the completion criterion now: every condition has evidence or an explicit 'could not determine'. Do not research past that.