We were instructed following the dismissal of an employee for gross misconduct. The severity of the conduct made it unlikely to be an isolated incident, but beyond the known facts we had nothing firm to go on. No confirmed pattern, no identified witnesses, no obvious lines of enquiry. We did not know whether other incidents had occurred at all.

The collected dataset ran to 100GB. After deduplication and processing, we were working with over 100,000 documents. We were searching for needles in a haystack that might contain no needles.

The old approach

Before generative AI functionality became available on eDisclosure platforms, an investigation of this kind would have required extensive human review. Keyword searching is a useful tool, but it depends on knowing what to search for. When you do not know the nature of the wrongdoing you are looking for, keywords are a blunt instrument. The alternative — broad human review of a large dataset — is slow and expensive. Junior lawyers and paralegals would work through documents, escalating material to senior lawyers for assessment. The senior lawyers, with the judgment and experience to recognise significance, would be working at one remove from the data.

What changed

Everlaw’s Deep Dive functionality allowed us to approach the dataset differently. Rather than searching for specific terms, senior lawyers were able to ask natural language questions based on concepts: patterns of behaviour, indications of concealment, acknowledgements of risk, language inconsistent with normal business conduct. Lines of enquiry could be developed and iterated in real time, with senior lawyers interacting directly with the data rather than waiting for documents to be promoted to them.

Material unconnected to the original incident was uncovered in hours. The same exercise, conducted by conventional means, would have taken weeks or months.

Why it matters

The value of generative AI in investigations is not simply speed, though speed matters. It is the ability to surface what you did not know to look for, and to put the lawyers best equipped to recognise significance in direct contact with the evidence from the outset.