The most interesting changes at RLLB were small ones. Across four days at LegalOps.com’s annual gathering, conversations about agents, outside counsel and AI kept returning to incremental decisions about the way legal work moves. A threshold changes, a playbook gets rewritten, a panel gets tighter, a category of work goes somewhere else. An agent is allowed to take another step before a lawyer intervenes.

None sounds like a new operating model. Together, they begin to resemble one. Litigation Ops Temp 29

Karen Helten, Salesforce’s Outside Counsel Senior Manager, and Camila Nagano, iFood’s Head of Legal and DPO, approached the shift from different sides of the same stage. Helten was looking outward at law firms: how work gets allocated, what invoice data can reveal and what clients should continue paying lawyers to do as technology absorbs more of the underlying work. Nagano was dealing with the question before some of that work ever leaves the company. At iFood, a contracting process can move a request from Slack through Salesforce data and Jira toward e-signature, with certain low-risk appendices proceeding without a human in the loop.

That distinction kept resurfacing. In a workshop on agentic AI, the questions weren't particularly futuristic. What can an agent access? What is it allowed to do? Where does its authority end? Who supervises it? On the surface, this is agent governance. Underneath it, someone is deciding which judgments can become rules—and which rules can execute.

The question extends outside the department. In another session, Deborah Haile of Uber, Daniel Neises of Snap, Jean Yang of Onit and Sam Bernstein of Davis Wright Tremaine took up whether clients should require, regulate or preclude law-firm use of AI. Whatever answer individual companies reach, the question itself places clients inside a decision that once largely belonged to the firm: how the legal work gets produced.

Outside counsel selection is moving in parallel. Rates, staffing, billing compliance, cost performance and predictability can increasingly influence which firms receive work. Technology is tightening the distance between selecting a provider, setting the economics and allocating the matter.

What connects these conversations is not AI so much as allocation: which work goes to a lawyer, which lawyer or firm gets it, what the client will pay for it and what can move without a lawyer at all. That was the undercurrent at RLLB.

Legal work has always been allocated. What is changing is how much of the judgment around those decisions can be defined in advance and embedded into the systems through which the work moves.

A lawyer deciding where one matter goes makes the call once. Put the criteria into a rule and the decision can reach the next hundred. Set an escalation threshold and the system knows when a lawyer needs to enter. Change a billing guideline and a decision about one task can extend across a panel.

The people making those decisions are changing, too. Legal operations, procurement, sourcing, finance, knowledge and technology professionals increasingly help choose providers, establish criteria, define workflows and write the instructions agents follow. They don't need to make the ultimate legal judgment to influence when, where and how that judgment is required.

This isn't a clean transfer of control from one group to another, or from people to machines. It is more diffuse. Decisions that once traveled together as part of a lawyer’s judgment are being separated into rules, thresholds, economics and permissions, often with different people responsible for each.

Most solve an immediate problem: move the contract faster, control the bill, improve consistency, clear the queue. What is different now is that the solution can remain behind. The next contract encounters the threshold. The next matter follows the rule. The next invoice meets the guideline. The next agent inherits the permission.

One adjustment doesn't amount to much. Hundreds begin to rewire how the work gets done.

The legal industry is still debating what AI will eventually do to legal work. At RLLB, people were already making many of the smaller decisions that will help determine the answer: where the work goes, who does it, what gets paid for and when a lawyer needs to be involved.

Increasingly, those decisions can be written once and executed again.