Petra Paskernak, Everlaw: New Research: Concerns Aside, AI is Delivering Results for In-House Legal

In-house AI savings claims are plausible, but a vendor survey cannot tell you whose savings they are

Everlaw, the e-discovery platform vendor, released a survey report on 5 November 2026, produced with Wakefield Research. It covers 300 in-house legal professionals. Of those, 86% said they use generative AI tools for legal work, and 92% said they see measurable cost savings. Everlaw says part of that saving comes from handling more work internally instead of passing it to outside counsel. Respondents still flagged accuracy and other risks. They also expect new approaches to the billable hour soon. ACEDS carried an extract of the article by Petra Pasternak. We have seen only that extract, not the full report.

How much weight should a buyer give the 86% and 92% figures?

Treat them as directional only. The extract does not define "using" or what counts as a measurable saving. It also does not say whether the savings were audited or simply reported by respondents.

The sentence structure leaves a gap. It is unclear whether 92% is a share of all 300 respondents or only of the 86% who use AI. Those readings give different pictures. Everlaw sells software to the same audience it surveyed. The vendor chose the questions, and the headline reads as a sales argument. None of this makes the numbers false. It means they are self-reported sentiment, not a measured benchmark.

Who actually benefits when work moves in-house?

Corporate legal departments benefit first. The savings described come partly from keeping work away from outside counsel, so a cost reduction for the client is a volume reduction for the firm.

That is the angle the extract leaves unstated. If in-house teams can do more themselves, outside firms will be pressed on what remains. The survey says billing approaches are about to change but gives no detail in the extract. We covered a Harvey event where attorneys said AI's effect on fees needs transparency with clients. The two threads point the same way. Expect clients to ask firms how AI shows up in an invoice.

What does this mean for e-discovery and review buyers?

Expect your corporate clients, or your own general counsel, to ask why AI-assisted work isn't cheaper. Be ready to answer with matter-level evidence, not survey percentages.

The Everlaw Chief Legal Officer, Gloria Lee, offered a telling example. She said AI helps with things the team "really could not do before", such as harvesting inventions for patent disclosures. That is new work, not cheaper old work. Savings and expanded scope are different outcomes, and a single percentage can blur them. Meanwhile the persistent accuracy worry is the one that matters in litigation. It argues for tools whose answers link back to the source document, so a reviewer can check them.

What should I ask my vendor before the next renewal?

Ask how they define and measure savings, on whose matters, and with what validation of accuracy. Then ask where privileged data is processed and whether any of it trains a model.

Also ask whether a quoted figure comes from customer self-reporting or from measured workflows. If the answer is a survey like this one, you have a marketing claim. Request a pilot on one of your own collections and compare time and cost against your current process.

Frequently asked questions

Is a sample of 300 enough?

It is modest for sweeping claims about in-house legal. The extract does not describe how respondents were selected, so be cautious about generalising.

Does the survey say how billing will change?

Only that new approaches to the billable hour are on the near horizon. The extract gives no specifics.

Is the research independent?

Wakefield Research ran it in partnership with Everlaw, which sells e-discovery software. Read it as vendor-commissioned.

Sources: ACEDS, carrying an extract of Petra Pasternak's article for Everlaw.