Please enjoy this blog co-authored by Aseet Patel, Founder and CEO, PatentOAResponse.com and Matti Neustadt, CEO, Founder, V4 Final.
On May 21, 2026, ILTA's "Chasing AI ROI" webinar asked the question that a lot of firms are quietly avoiding: after all the money spent on AI, how do we actually know there was a return on the investment (ROI)? Moderator Holly Hanna (Sr. Knowledge & Innovation Solutions Manager, Perkins Coie) led panelists Floor Blindenbach (CEO, Organizing4Innovation LLC), Krista Ford (Director, Knowledge, Research & Information Services, Steptoe & Johnson, PLLC), and Ron Friedmann (Independent Consultant), in a lively discussion.
A. Stopping the celebration of adoption.
Panelists nearly unanimously agreed that moving away from measuring adoption and toward measuring impact was critical. Most firms point to login counts--seats filled, prompts run, tokens--to justify the budget to management. But adoption is a vanity metric. Login data cannot tell the lawyer who used the tool to finalize a brief from the one who burned an hour fighting through bad prompting. Worse are the common heuristic biases: the sunk cost fallacy and endowment effect. The sunk-cost trap mandates using a tool simply because the money is already spent and the endowment effect creates an impression of value simply due to ownership. True ROI was never about justifying the purchase; it is about a measurable faster, cheaper, better outcome.
B. Defining what you are buying.
Part of the confusion is that AI is both a baseline technology and a paradigm shift requiring change management. That is partly why it cannot be scoped like a software license. Buying AI requires more than signing a contract; it is a strategic investment in the training and follow-on support. Any calculation of ROI must take this ramp-up time into consideration, as well as the intrinsic cost to make the change.
Defining how you will measure the change impact is a prerequisite to understanding what you are buying. Often, firms implementing AI have no measured baseline against which they can judge the impact of AI. Other times, they don’t even consider what they will measure before implementing a system and mandating its use. Vendors for these tools will insist on various ways in which they improve intangible ideas such as “efficiency” but must be pushed to understand exactly how the system can be used to measure efficiency rather than relying on anecdotes and succumbing to the endowment effect. Defining the data by which you will measure success helps mitigate underlying biases. Data is the antidote to the endowment effect.
C. Measuring impact, not usage.
Tokenmaxxing practices have made headlines in the mainstream media as a vanity metric that has landed some IT spend in hot water. In the legal ecosystem, the real metric is: where are the clear efficiency gains? Are your attorney super-users billing more time and onboarding more clients? Are you winning new work because of AI? Are you taking on matters that simply were not economically viable before?
One cautionary note about ROI: effectiveness beats speed, but a business model that solely relies on a billable hour model might miss that mark. Faster output is not effectiveness if it needs two hours of senior attorney review to fix confabulations or miscites. Measure the reduction in the total task lifecycle — including human review time— not just raw drafting speed. Moreover, consider measuring metrics like improved client outcomes in terms of shorter closure cycles in M&A deals, faster collaborations with key attorneys in the firm (less client downtime), or increased ease of use for complex systems like patent docketing.
D. Managing client expectations (and pricing).
Clients demand transparency and most clients want to know the final cost of legal services upfront. The billable hour works against lawyers' ability to set final fee expectations, but when legal fees are capped, the incentive flips. Every hour AI saves widens the margin instead of erasing revenue.
Ben Weinberger, author of The Death of BigLaw: Why the Legal Industry's Business Model Is Failing – and What Clients Must Do Next, puts it this way: "Law firms are investing heavily in technologies that reduce the number of billable hours needed to deliver legal services. AI isn't cannibalizing law firm revenue; it’s exposing a business model that cannibalizes itself whenever lawyers become more efficient."
Maximizing AI ROI while selling by the hour is difficult, but recognizing that client expectations actually go beyond pricing, whether hourly or fixed, is the paradigm shift. Clients are increasingly demanding to pay for outcomes rather than hours. Enterprise corporate clients are often not price sensitive (meaning they will not change law firms simply because rates go up) and expect valuable outcomes for the high fees paid. Transparency of how AI is used to drive better client outcomes can support both hourly rates and final invoices that boost firm profitability even while using traditional metrics.
Profitability is the other side of the pricing coin. Stop measuring revenue from a book of business and start measuring margins. ROI equals fixed-fee revenue minus the sum of internal labor cost and AI technology cost. The catch is that none of this sticks if your compensation system still rewards 2,500 annual billable hours. Law firms must restructure incentives to reward efficiency and margin realization, and effective AI use will naturally follow.
This challenge reflects a broader structural issue facing the legal industry. As Weinberger explains, firms are increasingly investing in technologies that improve efficiency while continuing to rely on pricing and compensation models built around the sale of lawyer time. Demonstrating a meaningful return on AI investment therefore requires firms to rethink not only how they measure success, but also how they price legal services and align incentives around value delivered to clients rather than time spent delivering it.
Conclusion
The ILTA panel saw agentic AI trimming overhead, clients potentially insourcing work they once sent out, and a hard question about training the next generation. Capacity reclaimed from rote work is a chance to invest in real skill-building — and in lawyer well-being, an ROI rarely counted. None of it happens by default; the real barrier is human inertia, not technology. And all of it rests on ethics and compliance.
Missed the initial session, Chasing AI ROI? Check it out here.
#KnowledgeManagementandSearch#KnowledgeManagement#Just-in-Time#ArtificialIntelligence#GenerativeAI#100Level#Adoption