Share this post:
Subscribe:
Your law firms are adopting AI, and many are genuinely saving time doing it. What almost nobody on either side of the relationship can yet show, at matter level, is how that saved time translates into staffing, turnaround or price.
In March 2026, Harvey announced an $11 billion valuation and said more than 100,000 lawyers across 1,300 organizations use its platform. It also said it partners with a majority of the AmLaw 100. In June 2026, Legora reported more than $150 million in annual recurring revenue and more than 1,000 customers across 50 markets.
That is real, and the direction of travel is not in doubt, but licenses bought are not the same as capability built.
That gap is becoming more consequential as clients ask firms to show the financial impact. The Financial Times reported in September that Wall Street financial institutions including Morgan Stanley and Citigroup are seeking alternative billing structures with major law firms as AI changes how legal work is delivered. Goldman Sachs has asked firms how much the technology is saving them and expects to share in those benefits.
The pressure raises a harder question for legal departments. Even if AI makes outside counsel more efficient, how do clients determine what those gains actually are?
The Core Question
Everyone agrees, AI is improving the efficiency of legal work, but almost nobody can say by how much. Firms are using AI and clients should share in the gains. But at a matter and task level, no one knows how to measure efficiency. Firms often don’t know how long a task actually took because their systems were designed to capture revenue, not measure operational efficiency. Law departments face a similar challenge. They have historically treated invoices as a hassle and invoice review as a compliance check rather than a source of insight.
Attempts to measure AI’s impact are complicated by uneven adoption. Many firms are still working to mature from individual ad hoc tool use to repeatable, measurable workflows. A lawyer using AI to summarize a document is making a real gain in efficiency, but that is a different thing from a firm that has built AI into a defined workflow, applied the same way across a team and measured as it goes. None of that is a criticism of lawyers or firms, it is the ordinary shape of enterprise technology adoption.
The result is that both sides end up relying on instinct rather than evidence, so the question of how much AI actually saved gets a shrug on one side and suspicion on the other.
Most Gains Are Absorbed or Obscured
Task level efficiency does not automatically translate to a lower invoice, because if a lawyer saves time on a research task, but the matter is staffed the same way, billed under the same rate card and recorded under the same narrative codes, the financial effect may never surface in a form the client can evaluate.
We know AI helps lawyers work faster. What is harder to see is where those gains end up, and much of it is likely absorbed into the way work already gets done or lost in how the work is reported back. From the client’s side the benefit is real but invisible.
Lower bills should not be the only measure of AI’s value to legal work, because AI may also contribute to faster turnaround, greater capacity, improved consistency, better predictability and lower risk. Measuring those benefits runs into the same problem as measuring efficiency, since clients rely on evidence to work out what they actually received.
Despite those challenges, the pressure on law departments to quantify what AI has gained them is only going to increase, as CFOs and procurement teams keep reading headlines about what AI is doing to the legal industry while their own legal spend holds steady or rises. They will question why nothing looks different if AI has changed everything. That answer needs evidence rather than assertion.
Establishing a Baseline Is Essential
Measuring the return on AI requires something to measure against, If you don’t know where you started, you can’t know how much has changed, which is the first step to giving credible answers to the people asking these questions.
Someone must build that starting point, and it is unlikely to be the firm. A firm that measures and reports its own speed gains is, in effect, making the case for its own rate reduction, so the incentive to do it isn’t there. Law departments are in a different position; they are the ones who want the answers and the ones who will act on it.
A spend assessment builds that starting point. It takes the invoice data you already own and answers a question it has rarely been used for: what does this task cost per unit and is that number moving? That is your baseline, and once it exists, AI’s effect on your bills stops being an argument and becomes a measurement.
In one 2026 Integreon assessment, approximately 15% of submitted outside counsel spend was identified as potentially recoverable before any rate negotiation, after an analysis at the invoice and guideline level.
Three Practical Steps
1 - Establish the baseline.
Starting by building a current state view of matter cost, phase and task mix, staffing, elapsed time, write-offs, and billing guideline exceptions. The goal is to understand where the work is being performed, by whom, and with what degree of variation. Without the baseline, it is difficult to determine whether changes in delivery are creating value.
2 - Agree what good looks like with your firms.
Ask firms to provide meaningful task narratives, explicit staffing assumptions and a sensible approach to transparency around material AI use. For repeatable work, work together to determine how the delivery model should influence scope, staffing and pricing. The objective is not simply to demand lower fees, but to create a shared understanding of how the work should be delivered.
3 - Define Value before negotiating it.
Decide what AI driven efficiency is intended to achieve: lower unit costs, faster delivery, greater capacity, better predictability, or some combination of these outcomes. Once the desired outcome is clear, it becomes possible to establish the right measures and manage performance against them. If value is not defined upfront, it cannot be measured or managed.
The question is no longer whether AI is changing legal work, it is whether clients and firms can show where the value appears. Legal departments that establish that baseline will enter the next round of outside counsel conversations with evidence rather than assumptions.
