Somewhere this week, a consultant is telling a board that AI can take 30% of the legal department’s budget. Down the hall, the general counsel is working out how to pay for an AI rollout that arrived with no money attached.
That gap between what boards are hearing and what legal teams are living with ran through the marquee session I moderated at last week’s Association of Corporate Counsel (ACC) Annual Meeting in Boston, with Shannon Klinger of Moderna, Carolyn Herzog of Elastic, and Hossein Nowbar of ServiceNow.
Michele Gorman of Law360 Pulse captured the conversation thoughtfully, and I’d encourage you to read her article, linked at the end of this post. Here I want to take the conversation a step further, drawing on what I hear from general counsel as they plan budgets, build their teams, and rethink how they use outside counsel.
Bring your own numbers
The 30% figure came from Shannon Klinger at Moderna, who told the room that consultants are pitching CEOs and boards on Legal as the last professional service AI hasn’t disrupted. Few GCs would accept that number, but a board might if the legal team has nothing better to offer.
Budget season makes this urgent. GCs are sitting down with their CFOs right now, and many are being asked to deliver AI without new money to do it. In our 2027 In-House Legal Budget Report, nearly every legal leader we surveyed said at least some of their AI initiatives arrive without a dedicated budget. ACC's 2026 Chief Legal Officers Survey tells the same story: CEOs increasingly expect their CLOs to lead on AI, and budget remains the biggest obstacle. So Legal is expected to deliver AI, fund it from existing budgets, and prove the return. All three, at once.
From my conversations at ACC, the GCs I see handling this well treat each AI project as a small experiment, with a stated hypothesis and a baseline taken before anything changes. It isn’t glamorous work, but it gives them a credible answer when the 30% figure comes up in the boardroom.
Hossein Nowbar of ServiceNow described another way through; his team uses AI agents to connect HR and legal systems that had never talked to each other, turning scattered data into a usable signal. Because HR benefits directly, HR pays for it. The money for legal AI doesn’t always have to come from legal’s own line.
Deciding where the legal work goes
Shannon Klinger put the stakes plainly in the Law360 piece: legal leaders need to decide “what work we want to keep internally to continue to develop our own talent versus what we want to send to outside counsel.” That is a talent decision as much as a sourcing one, and it will shape legal budgets for years. Too often, nobody chooses at all. The law firm reflex takes over, and overflow work goes to whichever firm handled the last matter.
The outside counsel picture shows why that matters. At a separate ACC session, “The AI Advantage: Resetting Expectations with Outside Counsel,” a law firm panelist said most of the AI guidelines their firm receives from clients prohibit or heavily restrict AI use, and few have been revisited since the early hallucination stories. A show of hands in the room suggested formal AI guidelines for outside counsel are still the exception. Meanwhile, in our 2026 In-House Legal AI Report, 92% of legal leaders said they expect lower rates when firms use AI, but most aren’t getting them.
The result is that many clients restrict AI in their guidelines while seeing none of the savings on their invoices. A short conversation at the start of each matter about what AI should touch, whose tools will be used, and how any savings will be shared would resolve much of that.
The larger opportunity is to revisit the work that needs a law firm in the first place. Much of what goes out the door, such as contract review, regulatory updates, and overflow on routine matters, doesn’t require a firm’s full rate structure. It can be handled by an in-house team using AI, by experienced lawyers brought in on a flexible basis, or by a combination of the two, freeing up outside counsel budget for the matters that truly need it.
AI adoption is a leadership job
The legal teams furthest along with AI didn’t wait for a mandate. Their leaders went first.
About three years ago, Shannon Klinger sat her leadership team down for a two-hour, hands-on session. Each person brought a real problem and worked through it with a generative AI tool, and most left with the start of a solution. Moderna’s legal team now uses AI to compare the terms in its standard licensing agreements against deals its peers have filed publicly, keeping its own terms in line with the market. It also uses AI to monitor compliance processes such as interactions with healthcare professionals. Carolyn Herzog’s team at Elastic followed a similar path, exploring potential use cases freely at first and then committing to a few focused cases, starting with an internal chatbot that answers routine NDA and policy questions.
Most teams haven’t invested that kind of time. In our legal AI report, nine in ten legal team members had received ten hours or less of AI training in the past year.
Governance belongs in the same conversation, because an overly restrictive policy can stall AI adoption as surely as a lack of interest. At “The AI Advantage” session, one in-house team described scrapping a first draft of AI guidelines that required pre-approval for nearly everything. The replacement runs two pages and rests on a few non-negotiables (such as no training on client data). A Women Influence & Power in Law (WIPL) conversation I moderated recently reached a similar view: start with enterprise tools, share only the data a task requires, keep a person reviewing the output, name the areas AI shouldn’t touch, and revise as the technology changes.
How junior lawyers will learn
I asked our panelists how junior lawyers will develop when much of the routine work that trained most of us is being automated. The panel treated it as an existential question for the profession, and it’s the one I hear most often from GCs once our conversation turns to talent. Junior attorneys feel it too. A Law360 Pulse survey this spring found many of them worried that AI will take over the work that has traditionally launched legal careers.
Hiring habits are shifting in ways that sharpen the problem. In our 2027 In-House Legal Budget Report, most in-house teams said they now ask whether AI can handle the work before they open a new role. That’s a sensible check, but if it leads to fewer entry-level hires, and first drafts increasingly come from a tool, junior lawyers lose much of the repetition that builds judgment.
The panel still found reasons for optimism. Lawyers are trained to ask good questions, and that skill carries directly into working with AI. Junior lawyers may spend less time on first drafts and more time with clients and the business. Getting there takes deliberate choices about how work is assigned. One practical step is to have junior lawyers review and mark up the AI’s first draft, explaining what they changed and why, so the tool replaces the typing but not the thinking. Fewer than half of the teams in our AI research have formal peer review of AI-generated work in place, so many have room to start. It also takes leaders who treat teaching as part of their role, and closer work with law schools and firms on how training changes.
Until that pipeline catches up, legal departments will be short on experienced lawyers who know the work well enough to spot a flawed AI output and who already know how to use the tools. That gap will be hard to close through permanent hiring alone. Seasoned, AI-ready lawyers brought in for a project, a spike in demand, or a technology rollout can give a team the capacity and expertise it needs. Working alongside junior lawyers, they can also show them how to question an AI draft and catch what it gets wrong, helping the in-house team develop its own bench.
Every legal team is at a different stage, and nobody has to solve all of this in one budget cycle. But the conversation about that 30% is happening in boardrooms right now, with or without your numbers. If you’re working through any of this, I’d be glad to compare notes.
Read Law360’s coverage of our panel: “CLOs Urge Teams To ‘Lean In’ On AI Experimentation”