Demand is rising while legal headcount and outside spend flatten. Robert Scott’s bet is that some legal judgment can be captured once and applied many times—raising a harder question about what clients are paying lawyers to do again and again.
Corporate legal departments have a math problem.
Demand is rising, particularly around regulatory compliance and cybersecurity, while the traditional release valves are tightening. The 2026 CLOC State of the Industry Report, based on the 2025 Harbor Law Department Survey of 135 law departments, found that just 32 percent expect attorney headcount to increase. Only 37% expect outside counsel spending to rise, down from 58% a year earlier. At the same time, 85% now have dedicated AI oversight or resources.
Legal teams are being asked to absorb more work without simply adding lawyers or sending the overflow outside. The question is whether AI merely allows that work to move faster or begins to change the economics of doing it.
Robert Scott has been working on one version of that problem for several years. A technology lawyer who has represented IT managed service providers since the early 2000s, Scott built Monjur around the recurring contract work those businesses generate. He says the platform launched in 2022 with 100 clients by year-end. In September 2024, the company reported reaching $2 million in annual recurring revenue; Scott now puts the client base at roughly 1,100.

Monjur is not itself a law firm. Scott describes a corporate entity that holds the technology business alongside an affiliated firm where the licensed attorneys work. Monjur’s website says legal services are provided under contract with Scott & Scott, LLP and its network of affiliated law firms, and that a Monjur subscriber also becomes a client of “the law firm.”
The more striking numbers came during our conversation. Scott says two lawyers handle the subscription services for those 1,100 clients, while other attorneys take disputes, M&A and above-scope matters. He puts subscription revenue at roughly $500,000 a month, split across two lawyers.
“With AI, you don’t need headcount to create scale,” Scott says.
That’s Scott’s claim, not proof the billing pyramid is collapsing—but it points at something bigger than a productivity gain.
Law firms have historically created leverage through people. Senior lawyers supervise work performed by more junior lawyers; capacity expands with headcount, and the economics depend in part on how efficiently work moves through the pyramid. AI introduces the possibility that some legal output can grow without lawyer headcount moving in lockstep.
Contracting makes the distinction easier to see.
Take a limitation-of-liability provision. A lawyer establishes the company’s preferred language but generally knows the fallback positions the client can accept if a counterparty objects. Those positions may remain in a playbook, prior agreements or the lawyer’s own experience. Yet every time the issue comes up, someone still has to retrieve that judgment and apply it.
Scott’s approach is to put more of those choices into the system itself. Monjur’s contract knowledge bases contain the preferred provision along with several lawyer-developed alternatives. The AI can then help a client navigate a routine negotiation within those boundaries, returning to the lawyer when the issue falls outside them.
As Scott puts it, the client can proceed without checking back with the attorney “because I already wrote those alternatives.”
He describes it as a shift from static templates to AI knowledge bases that can actively redline and negotiate, not just sit there.
A traditional contract records a lawyer’s judgment. An AI-enabled system can begin to apply it.
Legal practice has long bundled together two different kinds of work: making a decision and repeatedly applying it.
Legal departments already understand that distinction. Contract playbooks exist precisely because a company does not need to reconsider every acceptable liability cap, indemnity position or approval threshold from first principles. The underlying decision has already been made. The recurring work is applying it correctly and recognizing when an exception requires fresh judgment.
AI makes that separation more consequential because the playbook no longer has to sit passively on a server waiting for someone to interpret it. A system can begin to act on it.
That matters against the operating pressure CLOC is documenting. If only about a third of departments expect to add attorneys while workload continues to grow, the useful AI question is not which product can produce the fastest first draft. It is where the department continues to consume scarce lawyer time applying decisions it has already made.
Contracting is an obvious place to look, but hardly the only one. Approval thresholds, recurring policy questions, established risk positions and escalation rules all contain some version of the same problem. The opportunity is not to automate judgment indiscriminately. It is to distinguish work that requires a new decision from work that repeatedly consumes one already made.
That distinction also creates a harder conversation with outside counsel.
The 2026 Thomson Reuters AI in Professional Services Report found GenAI use rising sharply. Forty-one percent of law firms and 47% of corporate legal departments reported that their legal teams were using it. Across professional services, however, only 18% of respondents said their organizations tracked AI return on investment, while another 40% did not know whether ROI was being measured at all. Two-thirds of corporate respondents said they wanted their outside firms using AI.
The visibility gap is just as notable. Thomson Reuters found that 68% of corporate legal respondents did not know whether their outside firms were using AI on their matters. Half of legal professionals surveyed viewed AI as at least some threat to law-firm billing and revenue. One corporate legal respondent put the issue plainly: “Billing has remained the same as it did before.”
Scott frames the accountability question from the other direction. His objection is not to AI in legal work but to AI without a lawyer behind it. Unsupervised large language models, he argues, can leave small-business users unable to judge whether the work is sound or know who is responsible when it is not. His model, he says, uses technology to support an attorney-client relationship rather than substitute for one.
That does not mean every efficiency should appear as a discount on an invoice. Clients pay for more than the labor required to produce a document. They pay for expertise, accountability, commercial understanding and the lawyer’s ability to recognize when the ordinary answer is wrong.
But “Are you using AI?” is becoming a fairly weak question for a sophisticated legal department.
The better questions are what changed because the firm used it. Did turnaround improve? Did the scope of service expand? Did the staffing model change? Was work that once required junior-lawyer hours removed from the process? If the same result now requires materially less human labor, where does the value created by that efficiency go?
Scott dealt with that problem himself, dropping hourly billing for subscriptions in 2022. He sees AI as awkward inside a model closely tied to time because the technology is designed to compress the labor required to reach an outcome. In his words, the technology is “incongruent with the traditional business model.”
The billable hour has outlived plenty of obituaries, and Scott’s timeline may be more aggressive than the evidence supports. Clients are not simply buying hours, and law firms are not interchangeable production facilities. Judgment, reputation, institutional knowledge and responsibility still matter, particularly when the facts no longer fit the playbook.
Scott is nevertheless pushing the idea further. He says he is developing a prototype for autonomous contracting in which one party’s AI agent could negotiate with another using established positions and escalate only the points the systems cannot resolve.
“The humans will get involved only when the AIs can’t find a suitable match,” he says. Scott believes lengthy agreements could eventually navigate routine negotiation in seconds. “This is not science fiction,” he says. “This is something that will be happening inside of 18 months from now.”
Eighteen months is a forecast, not evidence. Commercial negotiations rarely reduce neatly to clause selection. Relationships matter. Leverage matters. Facts change. Often the exception matters more than the rule.
Autonomous contracting does not have to arrive on Scott’s schedule for the fundamental economics to matter now.
At the same time, firms and clients are adopting AI with surprisingly little visibility into the financial return—or, in many cases, even how the technology is being used.
Scott’s model puts those issues in the same frame. His bet is not simply that AI makes a lawyer faster. It is that some legal judgment can be captured once, bounded by lawyers and applied repeatedly, bringing the lawyer back when a genuinely new decision has to be made.
A practical place to start would be to look for work where the company has already made the underlying decision but continues paying people to make it operational each time.
Then ask what part actually requires a lawyer.
Setting the rule may. Handling the exception probably does. Taking responsibility for the judgment certainly can. Applying the same approved position for the 50th time is a different kind of work.
The same distinction belongs in conversations with outside counsel. Clients do not automatically own every dollar of efficiency a firm creates. But when the production model changes, they should be able to see what else changed: staffing, turnaround, scope, price or some combination of the four.
AI does not make legal judgment less valuable. It makes the cost of repeatedly applying that judgment harder to hide inside the cost of creating it.
That is the pressure point for legal services: not whether judgment still commands a premium, but how long repetition can command one too.
About Robert Scott
Robert Scott is managing partner of Scott & Scott, LLP and a founder of Monjur, an AI-enabled legal services platform serving IT managed service providers and other small businesses. A technology lawyer since the 1990s, he has spent more than two decades advising MSPs on contracts, disputes, M&A and legal issues. His current work focuses on subscription legal services, AI-assisted contracting and new delivery models.
