By Rick Carter, Chief Commercial Officer, LDM Global
Relativity renamed RelativityOne to Relativity aiR this month, and its 2026 pricing puts the AI tools inside the subscription. There will be plenty of commentary on the name. I want to answer the questions buyers should be asking instead, because our position holds whether you run your matters in Relativity, Reveal, Everlaw, Disco, or anything else.
Is most eDiscovery work still litigation?
No. According to Relativity, more than half of the new matters customers are creating on its platform are not for litigation at all. They are data breach response, internal investigations, and regulatory work.
That is the real story behind the rename. The platform is catching up to a shift in the work itself, and that shift changes what a review provider should be judged on. In a breach matter, the deliverable is not a coded review set. It is an accurate notification list, delivered by a date that does not move.
What does a managed review provider deliver when AI handles first-pass review?
A defensible result, not a completed queue. If first-pass review comes bundled with the platform, a provider cannot credibly sell you the first pass. What is left is everything that makes the result hold up when opposing counsel or a regulator asks how the work was done:
- A documented methodology
- A validation sample the client can stand behind
- Privilege calls made by someone qualified to make them
- Quality control that is measured, not asserted
- A named person who answers for the outcome
No platform sells that, bundled or otherwise.
Frankly, that is what clients were always paying for. The hours were just how it got billed.
Can AI handle data breach response review?
Partly. AI is genuinely strong at the identification pass and materially weaker at the judgment work that produces the notification list. I want to be careful not to over-claim here, because the honest answer is mixed.
Where AI earns its keep: surfacing PII and PHI patterns, clustering near-duplicates, and prioritizing the documents most likely to contain sensitive data. That work is faster than it was two years ago, and it will keep getting faster.
Where it falls short is entity resolution, which is where the actual deliverable lives.
Why is entity resolution the hardest part of breach notification?
Because it is judgment work, and a mistake in either direction is costly. Entity resolution means deciding that four name variants across three file formats are one human being, reconciling that person against an institution’s roster, and determining whether a partial record crosses a given state’s notification threshold.
Get it wrong and the client over-notifies, which is expensive, or under-notifies, which is worse. The statutory deadline does not move and accuracy is not negotiable, so the human validation layer does not compress at the same rate as the review itself.
Does AI reduce the cost of document review?
It reduces first-pass hours, but it changes the mix of work more than the value of it. I would not tell you AI leaves the economics of review untouched, and I would not tell you the advantage of a large, trained team disappears either.
The hours shift rather than vanish. There is less first-pass volume and more QC, exception handling, entity resolution, privilege, and second-level review. That is higher-skill work, so expect total hours to come down while the blended rate goes up.
And data volumes are not shrinking. Every efficiency gain in review has historically been absorbed by growth in the underlying data. I see nothing in this announcement that changes that.
Does surge review capacity still matter with AI-assisted review?
Yes, and arguably more than before. Breach and investigation timelines are set externally. Being able to stand up a large, trained team quickly is what lets a client say yes to a deadline they cannot control, and AI does not solve for that.
At LDM Global, that capacity is concrete: we can add 250 trained reviewers in a week on top of the team we already run. What those reviewers spend their time on is changing. The reason clients call us, a hard deadline and more data than their own people can handle, is not.
Does it matter which eDiscovery platform your review provider uses?
Not as much as the result they deliver on it. We work across Relativity, Reveal, Everlaw, Disco, and many others, because our clients do. Our position does not depend on what any platform is called this year or which features come in the subscription. The value was always the result, and the ability to scale to meet it.
What should you ask a document review provider about AI?
Ask how they make the result defensible, not how fast the AI is. These five questions separate a provider selling a result from one selling a queue:
- How do you validate what the AI surfaced, and can I see the sample?
- Who makes the privilege calls, and what qualifies them to make them?
- In a breach matter, who owns entity resolution, and how do you check it before the list goes out?
- How fast can you stand up a trained team when the deadline is set by a regulator, not by us?
- When someone asks how the work was done, whose name is on the answer?