Nuix adds audit-logged AI agent to Neo investigations platform

Nuix has launched an AI agent for its Neo investigations platform that records every document it reads and every query it runs as it works. The Nuix Neo AI Agent targets legal, regulatory and criminal matters, where evidence must survive peer review, board scrutiny or court challenge.

Auditability is the real selling point

Generative AI has moved quickly into legal technology, but defensibility has lagged behind capability. Investigators and litigators can use a model to surface patterns in a document set. They struggle to explain to a regulator or judge how it got there. That gap has kept many legal AI tools at the edge of high-stakes matters, used for triage but not for evidence.

Nuix is betting that a transparent agent closes the gap. Its pitch rests on provenance rather than raw model power.

Logs are created as the agent works

According to Nuix, the agent writes its activity record at the moment of execution, not after the fact. The log covers documents read, queries run, connections followed and outputs returned. Users can export the full conversation, including the reasoning path, for review. Each answer links back to source documents so claims can be checked against the originals.

The agent runs on the Nuix Neo Gen AI Framework, which connects supported large language models to data already processed in Neo. It can work across whole case inventories, call more than 20 platform tools and generate knowledge graph queries automatically. Nuix also says its processing engine is deterministic and reads more than 1,000 file types at binary level, including corrupted and encrypted files.

Administrators get spending caps, token monitoring, prompt guardrails and role-based access. Deployment is available in the cloud or on-premises, which matters to customers with data sovereignty constraints.

Rivals face a higher transparency bar

The release lands in a crowded field. E-discovery, investigations and compliance vendors are all adding generative and agentic features, and most now promise some form of citation or grounding. Nuix is pushing the argument one step further: that a log of the agent's own actions, not just a link to a source, should be standard for regulated work.

That framing plays to Nuix's installed base of government agencies, financial institutions, law firms and regulators. It also gives enterprise legal buyers and legal operations teams a sharper procurement question. Vendors that cannot show how an AI agent reached a result may find themselves excluded from matters likely to face scrutiny.