Lots of talk and money these days are being focused on AI in legal. And like much of the talk, a recent survey and report from Everlaw is a classic case of viewing the GenAI glass as half full. (Well, perhaps 49% full, see below.) That’s not to say Everlaw is wrong, but viewed critically, the results could just as easily be seen as a glass half empty. And that questioning is important in thinking about where legal is with GenAI, where it’s going, and how quickly.
Everlaw is a cloud-based e-discovery and data management platform. The survey report is entitled the 2026 Legal AI Adoption & Impact Report. The underlying survey is the fourth annual one and was conducted with the Association of Certified E-Discovery Specialists and the International Legal Technology Association.
They emailed legal professionals in May and June of this year and received 257 completed responses.
An Industry Standard?
Everlaw reports that 49% of those surveyed are currently using GenAI in their legal work. Everlaw concludes that “it’s clear this technology is quickly becoming the industry standard.”
But 49% does not necessarily make an industry standard. With all the talk about GenAI, the emphasis at most conferences and in the media, one could argue that the use should be higher. Certainly, those of us who go to conferences and write about the technology may be in our own echo chamber. But there is no mistaking the fact that over half of Everlaw’s respondents are not using the technology despite all the talk.
Everlaw notes that 21% presumably of the 49% who report using it are using AI multiple times a day. Nineteen percent report using it daily; 23% weekly; 3% monthly. The first two are increases from the previous year. Yet 34% say they use GenAI only occasionally. And for specific needs. Again, that’s hardly an industry standard.
Almost half of the respondents (46%) say they believe generative AI will soon be the standard across the practice of law. What’s not clear is whether this is 46% of the 49% that are using generative AI currently, or 46% of the total. Either way, that does not necessarily mean that generative AI will automaticlaly “become embedded across legal departments and practice areas,” as Everlaw says.
Everlaw also suggests that generative AI is becoming standard across organizations. Instead of being of more limited use by specialized teams or in isolated cases. But while 46% of the respondents say GenAI will soon be standard, the rest say not so fast. Of that group, Everlaw reports that 18% say it will be leveraged only for specific cases. Thirty percent say it will only be leveraged if it is proven reliable, and 3% say it will not be significantly leveraged. Granted, the share that believes it will not be leveraged at all is now very small. But that also does not mean that it’s going to be embedded across organizations at least in the near term.
A Frog on a Log
Everlaw cites the fact that, in addition to the 49% who are currently using AI, 42% say they are planning to use it in their legal work. Everlaw believes this further supports its industry standard conclusions. Forgive me, but I hardly think you can count those that are “planning to” use a tool but aren’t using it yet for much of anything. It’s like the old joke: three frogs are sitting on a log. One frog says, “I think I will jump into the water.” How many frogs are sitting on the log? The answer, of course, is three. Thinking about doing something is far different than actually doing it. The same applies here.
The Billing Conundrum
Only 19% of Everlaw’s respondents believe that GenAI has significantly altered conventional billing practices. While that’s an increase from the year before, that’s still less than one-fifth. Looking at the glass as half empty suggests that a lot of legal professionals have not yet figured out how to bill appropriately for GenAI work.
Everlaw also concludes that adding in the 31% that are planning to alter billing practices in the next year evidences a fundamental shift. But again, that could mean there are 31% of the frogs on the log that are just thinking, not doing. And over half of the respondents agree that GenAI will “weaken” the conventional billable hour model is a far cry from saying that half of those surveyed have changed their billing practices away from the billable hour model or will do so in some fundamental ways.
For those who are using GenAI, 17% believe that it saves them 5 to 10 hours a week, which, of course, is a lot. But 21% are not sure it saves time. And you have to ask what time is being saved: billable time or non-billable time? It’s great if you’re saving 10 hours of administrative work. It’s not so good for the bottom line if you’re cutting 10 hours of billable time and not replacing that revenue somehow. Which brings us back to the need to look harder at the billable hour model.
Agentic AI? Not Hardly
Everlaw also looked at the use of agentic AI. Despite all the hoopla, only 3% of those surveyed are currently using AI agents. As in close to zero. When asked where agentic AI could have an impact in the future, the responses were all over the board. The highest percentage (42%) focused on the possibility of synthesizing fragmented data. I’m not sure what that means or how that is different than what GenAI can do apart from agents. It’s indeed possible that the statistics could reflect a lack of understanding of GenAI and agentic AI.
Some Alarming Stats
But here’s a really alarming finding for which it’s hard to see a half-full glass. Only 34% of those surveyed say they are integrating GenAI organization-wide. The rest are basically allowing one-off strategies for partners. Or for specific workflows or practice groups. Eighteen percent are still in the testing and/or piloting phase. 8% have no formal strategy whatsoever. That means little institutional control.
And we see this gap reflected in another stat. While 48% say they have written policies in place governing AI use, over half either have no policy (12%), are still planning to have one someday (21%, aka don’t have one), or have one only for employees (20%, as in it doesn’t apply to partners). Everlaw rightly calls the one-third without a policy “a striking statistic given that nearly half of the respondents report actively using generative AI in their work.” Yep.
Putting all those statistics together, we can see that governance is still spotty. Two-thirds aren’t integrating AI organization-wide. And one-third have no formal policies, which means that a good number of people in legal organizations are perfectly capable of doing exactly what they want with GenAI when they want it. That’s a dangerous combination.
And the Future?
And here’s another alarming statistic. When asked what AI proficiency is necessary for future development, 41% focused on output verification. That’s low, considering the issues with hallucination. It perhaps reflects a misunderstanding of the tool itself. Also, surprisingly, 14% say that no AI skills are even necessary.
Here are some more statistics that are difficult to view with rose-colored glasses:
– 31% say that AI will create a foundational gap in skills for younger lawyers;
– 25% say that with the use of AI, senior leadership will have difficulty assessing the capabilities of younger attorneys; and
– 25% say firms will hire fewer junior attorneys due to AI.
Added together, 81% of the respondents see some serious problems down the road, particularly with junior attorneys and their skill sets. Everlaw puts it well: “Firms will need to rethink how they build foundational skills and create opportunities for hands-on development for those junior attorneys.”
The Bottom Line
All of which is not to say that Everlaw’s conclusions are wrong. It is to say that some of the conclusions may be optimistic and there is a need to look at them critically.
The survey responses may indeed represent that there is a fundamental shift in how work is getting done and that AI is the next defining force in legal work. But the statistics still show that a large segment of the industry is not doing much when it comes to GenAI use. A lot of respondents are still just planning.
Some findings could reflect an industry that is still slow in understanding the potential impact GenAI may have on the billable hour model. There is little use or understanding of agentic AI. The need for governance across organizations is still not well appreciated. And there are significant concerns about future younger lawyer development.
So, it’s great that Everlaw has conducted the survey. It’s helpful, but like any statistics, we have to look at them critically and not jump to rosy conclusions.
Let’s look at the glass as half full, yes. But it’s also important to look at it as half empty as well. Particularly with a technology that is as disruptive as AI. There still could be three frogs on the log despite all the numbers.
Stephen Embry is a lawyer, speaker, blogger, and writer. He publishes TechLaw Crossroads, a blog devoted to the examination of the tension between technology, the law, and the practice of law.
The post Everlaw’s 2026 AI Report: A Glass Half Full Or Half Empty? appeared first on Above the Law.