Glossary / Agents & Automation

Deep Research

What vendors mean: AI that does hours of research for you. What it actually means: an agent that runs many searches over many minutes and writes a cited report — powerful, and still capable of confidently citing sources that don't say what it claims.

Agents & Automation

The term

“Deep research” names a class of agentic feature — now offered by every major model provider — where, instead of answering instantly, the AI plans a research task, runs dozens of searches, reads sources, and returns a long, structured, footnoted report. It trades speed for depth.

Why it matters to you

For market scans, competitive landscapes, and literature reviews, this genuinely compresses a day of analyst work into minutes, and the citations make it auditable. That’s a real gain for a lean team.

The failure mode is subtle. The report is fluent, formatted, and footnoted, which makes it feel more trustworthy than it is. A citation proves a source exists — not that it supports the sentence attached to it.

What that means in practice: Use deep research to get 80% of the way in 5% of the time, then spot-check the citations that matter before you act on them. It’s a brilliant first-draft analyst and a dangerous final authority. Deciding which claims to verify is the judgment you still own.

Get the next Brief

One operator. Every other Wednesday.

Plus the AI Glossary and the Failure Museum.
Real names. Real numbers. Honest analysis.