An AI first pass at responsiveness, and an agent that shows its work#
This release puts an AI first-pass responsiveness review inside the matter, and gives the Matter Agent live progress, charts, session sharing and a full per-session activity log — so reviewer time lands where judgment is actually required.
These updates build on Rational Review’s AI-powered eDiscovery platform, which brings document review, analytics and production into one workspace. Review Assist and the enhanced Matter Agent help teams focus reviewer attention where it is needed, collaborate on analysis and maintain a clear record of AI activity within each matter.
AI Review Assist
An AI first pass at responsiveness#
Review Assist performs a first-pass responsiveness review inside the matter. You describe relevance once — by uploading a case file or relevancy criteria and letting the platform extract them, by working through a guided chat, or by filling in the criteria form — and the platform codes the population from that description.
Every document comes back with a responsiveness decision, a confidence score, the criteria it matched, and a plain-language rationale, streaming into a live results table as the run progresses. The documents the platform cannot call confidently, and those where it disagrees with your team’s earlier coding, route automatically into a Needs Human Review queue.
Iteration results. Every document carries a decision, an evidence strength, the reason in plain language, and the passages it was drawn from.
- Describe relevance three ways — upload a case file or relevancy criteria, work through a guided chat, or complete the criteria form.
- Decision, confidence score, matched criteria and a written rationale for every document.
- Low-confidence calls and disagreements with prior human coding route to a Needs Human Review queue.
- Reviewer reasoning captured on those documents feeds the next iteration, so accuracy improves run over run rather than starting fresh.
- Every criteria change, run, resolution and export is written to the matter’s audit record.
Why it matters
Review effort goes where judgment is needed. Rather than every document getting the same first look, your team concentrates on the calls the platform could not make confidently and the places it disagrees with earlier coding — and the record of how the review was conducted is built as you go, not reconstructed afterwards.
Review Assist covers responsiveness review and is enabled per matter. Speak to your account team about turning it on.
Matter Agent
An agent that shows its work#
The work on the Matter Agent in this release is about oversight: seeing what the agent is about to do, watching it work, and being able to account for it afterwards.
Long-running work now reports live progress in the chat instead of appearing to stall, replies can return charts as well as text, and a session can be shared with named colleagues so an analysis is not locked to one person’s window. Behind all of it, every operation the agent performs is recorded.
- Live progress for long-running work.
- Charts as well as text in replies.
- Share a session with named colleagues, and manage who has access.
- Session owners can open a full activity log of every operation the agent performed, alongside usage metrics for that session.
Why it matters
“What did the AI actually do on my matter?” now has an answer you can open and read — every operation the agent performed on a question, in order, for the session owner to check.
Matter Agent is enabled per user. Speak to your account team about turning it on for your team.
On the Horizon
Coming next#
Spam Detection
Identifying junk — marketing and social traffic, system logs, phishing, boilerplate patterns — before it is hosted, processed and reviewed, with a per-matter domain whitelist so key senders are never swept up. Nothing is removed automatically: candidates are presented as a reviewable list with the reason each was flagged. In development; talk to your account team if you would like to be part of the early conversations.
