Legal departments have spent years measuring law firms after the work is assigned. AI is bringing that accumulated performance record forward, to the point where the next assignment is made.
For years, legal departments have built systems that explain outside counsel after the fact. E-billing exposed invoice detail; matter management created histories; rate benchmarks sharpened negotiations; billing guidelines and panels imposed discipline. Yet the most consequential choice sat upstream of almost all of it: who gets the work.
At Running Legal Like a Business (RLLB), “Stop Managing in Pieces: Architecting an AI-Driven Outside Counsel Program” examined what happens when the performance record generated downstream begins to reach that earlier decision. Panelists included Jennifer McIver, director of legal operations and industry insights at Wolters Kluwer; Daniel Decker, associate director of technical product management at Wolters Kluwer; Carol Bouw, head of global legal operations at Cargill and Arun Mistry, formerly of Starbucks.
Bouw recalled arriving in legal operations from a role supporting shared-services executives and being struck by the abundance of data inside the department—and how little the team recognized what it had. Matter histories and invoices sat alongside fee arrangements, billing guidelines and records of where work went. Legal was data rich, but the intelligence remained fragmented across the processes that produced it.
McIver’s framework brought those records together around five dimensions of outside counsel performance: billing compliance, cost performance, staffing leverage, timekeeper continuity and rates relative to market. Read separately, each answers a familiar operational question. Read together, they begin to describe the economics, performance and durability of the relationship.
A rate carries different meaning alongside the staffing behind it; staffing looks different against matter complexity; budget performance becomes more revealing across a portfolio than on a single engagement. Timekeeper continuity can show whether the institutional knowledge attributed to a firm reaches beyond one or two partners. The individual measures, viewed in context, begin to form a performance history.
AI extends that analysis into information legal departments have struggled to use at scale. McIver pointed to invoice narratives as one example. UTBMS codes provide useful structure, but the code attached to a line item may not fully describe the work performed and may have been selected by someone removed from the work itself. AI can examine the narrative alongside the code, billing guidelines and matter information, bringing previously cumbersome evidence into the analysis.
Bouw’s experience illustrates what happens when that intelligence reaches the allocation decision.
At a previous company, her team classified work by risk, complexity and frequency, developing a clearer view of what was entering the department and where it was being sent. In real estate, the analysis surfaced relatively low-risk work going to a mid-tier law firm because the internal team lacked capacity, even though the same in-house lawyer was already using a smaller local firm for other matters.
The issue was not poor performance by the larger firm. Better visibility simply made the existing allocation worth reconsidering.
Bouw’s team asked whether the smaller firm could handle more of the appropriate work and shifted it gradually. The change reduced annual outside counsel expense in real estate by roughly $80,000. Applying the same discipline elsewhere ultimately surfaced more than $500,000 in savings, according to Bouw.
What changed was not the panel of firms so much as the legal department’s ability to see how work was being allocated. It had enough information about the work, its risk and its historical sourcing to revisit choices that otherwise might have continued largely by default.
That pushes outside counsel intelligence into a different part of the operating model. Much of the technology built around outside counsel comes into play once the engagement decision has been made: the matter opens, hours accumulate, invoices arrive and legal operations measures, reviews, benchmarks and negotiates what follows. AI creates the possibility of carrying what those systems have learned into the decision that precedes the spend.
Decker’s discussion of cost performance sharpened the economics. The relevant measure is not simply whether a firm is cheaper, but whether its performance is predictable and consistent across a body of work—whether budgets hold, scope remains controlled and the work is staffed appropriately. A premium rate can coexist with disciplined execution; an attractive rate can lose much of its appeal when matters repeatedly drift beyond scope.
Rate, in that context, becomes one element of a longer performance history rather than a shorthand for value.
Sophisticated legal work contains genuine uncertainty, and no historical record eliminates the variables of a new matter. But repeated experience also creates information about staffing, effort and cost. Increasingly, clients have histories of their own against which to examine what firms know about the work and how confidently they can price it.
Continuity presents another dimension of the same problem. Bouw described the value attached to a deeply trusted senior partner whose knowledge of the company had accumulated over many years. Her department’s response was not to diminish that relationship but to preserve what made it valuable: involve the in-house lawyers in a succession conversation, develop the associates beneath that partner and begin placing work at the appropriate level. Institutional knowledge concentrated in one person may be valuable, but it is also fragile.
This is also where the limits of automation become clearer. McIver was explicit that the intelligence should inform the attorney rather than dictate which firm receives the matter. Outcomes, judgment, business knowledge and trust remain central, particularly when the matter at hand bears little resemblance to what came before. The change is in how much of the legal department’s own operating history can be present when that judgment is exercised.
For general counsel, outside counsel performance becomes less episodic and more cumulative. For legal operations, data once used chiefly to explain and control spend after the fact can begin influencing where work belongs before the commitment is made. For law firms, the implications extend beyond an annual review or the next rate negotiation: how today’s matter is staffed, budgeted and handled becomes part of the evidence surrounding tomorrow’s assignment.
The next matter will still require judgment. Increasingly, that judgment arrives with a longer memory.