In-House Innovation Report from Everlaw: Artificial Intelligence Trends

In-house teams say AI is shrinking their outside-counsel bill, but the evidence is self-reported

Everlaw released its In-House Innovation Report in October 2026, a survey of 300 in-house legal professionals run with Wakefield Research. eDiscovery Today covered it. The headline figure is that 86% of respondents use generative AI for legal work, and 37% use it extensively. Some 54% say they have already cut the work they send to outside counsel, and another 44% plan to start next year. Savings were reported by 92%, but only 13% call them significant, against 56% moderate and 23% slight. Separately, 81% expect most legal work to be priced by outcome rather than hours, and 47% expect that on the shorter of the two timelines the survey offered. The top worry, at 80%, is accuracy and alignment. Another 72% fear non-lawyers will lean on AI for error-prone legal advice, and 63% fear human skills will erode.

Can I trust the 86% adoption figure?

Treat it as directional only. The coverage doesn't define "using GenAI for legal work," and the survey was commissioned by a vendor that sells into this market.

Everlaw ran the research, and the outlet that reported it discloses that Everlaw sponsors it. That doesn't make the numbers wrong. But "using" can mean anything from drafting an email to running a review workflow. The coverage gives no sample-selection detail beyond the count of 300. Only 37% say they use it extensively, and that group is much closer to real production use.

Are the cost savings as big as 92% suggests?

No. Most respondents describe their savings as moderate, 56%, or slight, 23%. Only 13% call them significant, and the coverage reports no currency amounts at all.

"Measurable" is the respondent's own judgement, not an audited result. The coverage also doesn't say whether savings came from less work going out or from lower rates on work that still goes out. The author reads the results as outside firms passing some savings along, but that is his inference, not a survey finding. The 54% who have already reduced outside work is the more interesting number for discovery buyers. First-pass review is a typical task to pull in-house, though the coverage doesn't say which tasks moved.

What does the accuracy concern mean for evidence platforms?

It means buyers will judge AI on whether each output can be checked against a source document. With 80% naming accuracy as their top concern, fluent answers without citations will struggle.

The survey doesn't ask what respondents would accept as verification, and that gap matters. In litigation the practical test is whether an attorney can click from an answer to the exhibit behind it and defend it in front of a court. This fits themes we have covered before: insurers struggling to price law-firm AI risk, and the argument for smaller, sovereign models. All of them point to provenance and control as what makes AI usable on privileged material.

What should I ask my vendor or outside counsel?

Ask who measured the savings, against what baseline, and over what period. Then ask how AI-assisted work will be priced, since 81% of this group expect outcome-based billing.

For a discovery vendor, ask whether every AI answer links to its source exhibit, where the data is processed, and whether privileged material ever trains a model. For outside counsel, ask whether efficiency gains show up in your invoice or only in their margin. If your peers are already moving work in-house, as 54% say, you have leverage to ask for fixed fees.

Frequently asked questions

Who ran the survey?

Everlaw ran it in partnership with Wakefield Research, surveying 300 in-house legal professionals.

How many respondents plan to cut outside-counsel work?

54% say they already have, and 44% say they intend to start within the next year.

Does the survey prove AI reduces legal spend?

No. It records perceptions: 92% report savings, but most describe them as moderate or slight, and no figures in currency are given.

Sources: eDiscovery Today (Doug Austin), which in turn drew on Everlaw's In-House Innovation Report and a related Everlaw blog post by Petra Pasternak.