An M&A partner with a full book of business, strong precedent banks, and diligence work already automated or outsourced may not feel much urgency around GenAI. The current system works.
A senior associate turning messy information into something a partner can review will see value much faster — GenAI helps them structure documents, draft first passes, and prepare sharper questions before review.
A partner whose pipeline has softened may be more open still — not because GenAI has become more useful in the abstract, but because pressure around business development makes experimentation feel worthwhile.
With such an uneven landscape, measuring GenAI ROI in law firms can feel near impossible. Should firms simply treat it as baseline infrastructure and move on?
Probably not. Because understanding the data will also sharpen your adoption efforts. The real question is not whether someone understands GenAI — it is whether their current work gives them a reason to change. A simple comparison between "users" and "non-users" hides more than it reveals.
Law firm performance data is already incomplete
Before GenAI enters the picture, performance data in a law firm is already messy. Firms know the financial endpoints — hours, rates, billings, collections, utilization, realization. These metrics matter.
But much of the effort between the work done and the client paid is harder to see. Were all hours honestly booked? How long did it take to get from a messy first draft to something usable? How much senior attention was required before the work could move forward? How much rework was avoided?
Those moments are often where GenAI creates value — and they do not appear neatly in the data. GenAI adds another layer of ambiguity unless firms segment carefully.
Why GenAI ROI in law firms depends on user segmentation
It is tempting to divide people into users and non-users. It feels clean and gives leadership a simple adoption number. But the truth is more nuanced.
For ROI analysis, five groups matter: Non-AI-ers, Non-users, Ad-hoc users, Frequent users, and Daily users. Each tells a different story. Lumping them together distorts the data and leads to the wrong adoption strategy.
Non-AI-ers: the invisible high bar
Non-AI-ers have simply not seen a reason to engage with GenAI yet. What makes this group analytically tricky is that they are often among the strongest performers in the firm. Their output sets a high baseline — one that average attorneys using GenAI may still struggle to match. That can make GenAI look less impactful than it actually is, because the comparison group is quietly exceptional.
Non-users and ad-hoc users: stuck after a bad first impression
Non-users looked at GenAI, tried it briefly, and stopped. Ad-hoc users went a little further — a summary here, a first draft there — but never made it a habit. In practice, these two groups share the same problem: they are not fluent enough to get reliable results, which cements their belief that the tool has limited usefulness. The tool did not fail them. The onboarding did.
Frequent users: seeing value, not yet changing the work
Frequent users return to GenAI because something about it consistently helps — faster drafts, better-structured first passes, less effort on routine tasks. The value is real. But they are still doing the same work, just with less friction. ROI at this stage tends to be incremental.
Daily users: where work actually changes
Daily users are the most interesting group. They are not just using GenAI more often — they are working differently.
Where others use it as a faster horse, daily users are building the car. Work that once took weeks now takes hours. That starts a different conversation — about billing, about rates, about what a matter actually requires. That is where the most significant ROI lives. But this group is often small and easy to miss, partly because they are still experimenting heavily and their output can be uneven.
ROI analysis should shape adoption strategy
Segmenting users is not just about producing a more accurate ROI report. It is about knowing where to invest attention next.
Non-AI-ers need relevant examples, not generic training. Non-users and ad-hoc users need help overcoming early negative experiences — that is a heavier lift than most firms anticipate. Frequent users are ready to go deeper; moving them toward daily use is likely the highest-leverage opportunity available. Daily users need help identifying and scaling what already works, so the rest of the firm can follow.
The better question
For firms trying to turn GenAI investment into returns, usage data is only the beginning. The more useful questions are:
- Who has not tried it, and why?
- Who tried it but stopped?
- Who uses it occasionally but has not found a repeatable use case?
- Who returns to it because it reliably helps?
- Who is beginning to work differently because of it?
These distinctions change the interpretation of the data. They explain why ROI appears uneven, reveal which comparisons are fair, and prevent firms from treating adoption as a single linear journey when it is actually a set of different starting points, habits, and incentives.
The question is not simply whether GenAI correlates with better performance today. The better question is where GenAI is changing the conditions that make better performance more likely tomorrow.
GenAI ROI is not only a measurement problem. It is a segmentation problem — and a support problem. Because the journey matters as much as the destination.
For more, listen also to the ILTA round table on Chasing AI ROI.