Nuix launches Neo AI Agent with source-linked answers and an audit trail

Nuix launched the Nuix Neo AI Agent on October 22, adding a conversational investigation capability that links answers to source material and records the steps used to produce them. The Sydney announcement targets enterprises, legal teams, auditors, regulators and law enforcement working with large collections of information. Users can ask questions in ordinary language about documents, images and connections among people, accounts, communications and transactions.

The agent can draw on more than 20 tools within Nuix Neo. Its capability overview describes combining search, item retrieval and knowledge-graph traversal within a conversation. When a graph matching the current case name is available, the agent can inspect its structure and generate queries suited to that graph. This brings document evidence and relationship analysis into the same answer, while reducing the amount of query syntax an analyst must write.

The company puts the record of that work alongside the answer itself. Responses point back to source documents, and investigators can export the full conversation with the steps taken and findings. That gives another professional a record to examine when reviewing how an answer was reached. The launch announcement describes this as supporting explanations to courts and boards, with the underlying material available for verification.

Access and spending controls apply around the investigation. Nuix says role-based permissions are enforced across the case inventory, and its overview limits searches across cases to those the user can access. Administrators can set usage limits by cost, calls or tokens, organized by user, model, case or service. The documentation also describes a separate moderation model checking prompts and outputs against an organization’s standards, with blocking or audit actions available under the policies.

The agent operates on data prepared within the existing Nuix Neo platform. That platform brings together emails and attachments, documents, chats, spreadsheets, images, audio, video and forensic files. Nuix describes a processing foundation that can be reused across legal, privacy and investigation work. Its orchestration layer coordinates tasks and models with dependencies, routing and exception handling, while maintaining an audit record through the pipeline.

The launch follows Nuix’s October 8 announcement of the Neo Gen AI Framework, the governed infrastructure connecting a customer’s chosen model to the platform’s curated data. That earlier announcement had already made the agent available to Early Adopter participants through the Responsible Innovation Framework. Customers can choose a supported LLM and provider, while the framework handles governance, token metering and context management. The October 22 announcement develops the agent’s capabilities on that foundation.

The conversational workflow retains context during a session, so follow-up questions can build on previous findings. The overview also describes writing results into the case through item tags and custom metadata fields, making annotations searchable and exportable. Depending on the question, the agent can present a table, timeline, organizational chart, flowchart or narrative response. Those features keep findings connected to the underlying items after the initial exchange.

Nuix distinguishes this exploratory workflow from Neo Automate BYO AI, which applies a defined prompt to each item in a selected dataset and can produce structured reporting. The agent instead selects relevant items through tool calls to investigate open-ended questions and relationships. Both use the governed framework. The capability document assigns verification of outputs to qualified personnel before reliance, leaving professional judgment with the investigators, analysts and legal reviewers using the system.