Illustrative case study. This follows the Meridian demonstration on the NERVE website. It is not a customer deployment, and it claims no measured savings or operational results.

One question, three kinds of source

Someone needs to find everything held about the Meridian archive review. It is spread across a mailbox export, a document archive and a CSV import. Today that means running the same query three times and squinting at three differently shaped answers. Worse, by the time the results are pasted into a summary, nobody can say reliably where any of it came from.

Preserve first, then transform

In this workflow an operator brings the approved sources into NERVE along with how they were acquired and how they must be handled. The original bytes are kept before anything is extracted, so there is always something to go back to. Supported content is then turned into typed records that retain the detail specific to each source and the history of the processing run that produced them.

That common model lets the team search for Meridian once, across all three sources. It deliberately does not merge them. The message, the document passage and the CSV row stay separate pieces of information, because they are.

What the demonstration actually shows

The public Explorer example returns a message about a Friday review, a checklist inside a document, and a structured record where project equals Meridian. Each hit comes with its locator: a position in the message body, a page and paragraph, a row and column. The reviewer sees where a match sits rather than receiving an unattributed extract and being asked to trust it.

Access and analysis are different jobs

NERVE works out the authorised search boundary before it produces results. If you are evaluating this, include a user who is denied one dataset and then check the results, the counts and the facets. Counts and facets are where disclosure usually leaks. That is a test to run, not a result this case study is claiming.

Where CORTEX is involved, the analyst uses that information to build and review an assessment. NERVE supplies source material and evidence references; CORTEX provides the intelligence lifecycle. A shared project name or a repeated identifier is a reason to look, not a conclusion.

How to judge an evaluation

Agree what success looks like before you import anything. Can every accepted original be retrieved? Does each displayed match resolve back to its locator? Do access restrictions hold for a user who should not see one source? Does an earlier evidence reference still resolve after a new processing run? Record the unsupported formats and the failures as carefully as the successes.

The question worth testing is a plain one: can your team find approved information across supported sources and explain the chain behind every answer? Whether that saves time, and whether it scales to your volumes, only your own workload can tell you.

Try the Meridian Explorer demonstration

Read how NERVE handles source provenance

View the NERVE product

Plan an evaluation with your sources

Explore the platforms: nerve.aic.io and cortex.aic.io.