AI Modes

The AI Assistant has three modes, selectable from the chat input. Each controls how the AI investigates your question.

AgentResearch (Legacy)Deep Research (Legacy)
Best forMatter-bound analysis, case-law validation, anything you'd show a partnerFast online lookups — statutes, regulations, quick case scansLong-form narrative reports
OutputStructured legal memorandum with inline citations linked to sourceChat response with source linksLong-form chat response
SpeedMedium — investigation shown in real timeFastSlow
Default?YesOpt-in toggleOpt-in toggle

If in doubt, use Agent. It's designed for the full legal workflow — plans the investigation, reads your matter documents, validates case law, and returns a cited memo.

How to pick a mode

  1. Working inside a Matter with documents attached, or asking anything that touches case law?Agent. Built for the legal workflow. Plans an investigation, reads your documents, searches case law, and returns a cited memo.
  2. Just want a fast online lookup — a statute, a regulation, a quick "what cases address X" scan — with no memo?Research.
  3. Specifically want a long-form narrative report, and willing to wait?Deep Research. For most research needs, Agent now produces better-cited, faster output — try Agent first.

Agent

Flagship mode and the new default. Runs a structured investigation with inline citations that link back to source. See the full Agent Mode page for the complete breakdown.

Use Agent for:

  • Matter-bound analysis ("summarize the contract risks in this MSA")
  • Case-law validation ("pull and verify the controlling cases")
  • Any cite-ready answer you'd show a partner

Research (Legacy)

A single-pass lookup that taps online legal resources — federal and state courts, statutes, regulations — and the Irys Database of templates and secondary sources. Fast, lightweight, good for quick checks.

Use Research for:

  • "Is there a Texas statute on executory contracts?"
  • "Find cases about the implied warranty of habitability in New York."
  • Quick template or regulation pulls where a short answer is enough.

Research is a single-pass retrieval — it won't plan, pursue leads, or produce a structured memo. For anything you need to cite with confidence, Agent produces more reliable citations.

Deep Research (Legacy)

A longer, more comprehensive narrative report designed for memo-quality output where depth matters more than speed.

Use Deep Research for:

  • Long-form narrative reports you're willing to wait for.
  • Existing workflows you've built around it.

For most memo-quality queries, Agent now produces cleaner, better-cited, faster output. See the dedicated Deep Research page for more.

Next steps