Large firms are hiring for AI roles faster than they can find the people, and most of the roles are new. I hear the same thing in nearly every firm I work with. There are open requisitions for AI engineers, knowledge engineers and innovation leads, and the candidates who fit are rare. Firms have the budget for these roles, and they still can't fill them.
The money has already moved. The 2026 Report on the State of the US Legal Market from the Thomson Reuters Institute and Georgetown Law found that the average firm's technology spending grew 9.7% in 2025 and its spending on knowledge management tools grew 10.5%, which coverage of the report described as likely the fastest real growth the legal industry has experienced. Lawyer compensation rose 8.2% over the same period. Firms are paying for the tools and paying for the lawyers. The layer of people who connect the two is much harder to buy.
Why the usual hiring plan falls short
The instinct is to hire for AI the way a firm has always hired for technology: write a job description, look for the right certifications, and place the new hire in the IT org chart. That approach works for running a document management system or a network. It works poorly for AI, because the work that matters most sits between the law and the software, and very few people have spent their careers in both.
The entry level of these jobs is also changing quickly. The AI agents we have today are the most basic they will ever be, and a graduate with a master's degree is already competing with one for routine analysis and coding. A firm that fills its AI team with junior hires to do work an agent can already do will find itself staffed for the wrong job within a couple of years. The people worth competing for are the ones who can direct agents, check what they produce, and design the systems they run in.
The roles firms are trying to fill
Titles vary from firm to firm. The work falls into four groups.
People who capture what lawyers know. The most valuable thing a firm can do with AI right now is capture the tacit knowledge of its lawyers: the positions they took, why they took them, and the judgment that never makes it into a precedent bank. Much of that knowledge already shows up in the signals lawyers leave behind in email, drafts, time entries and matter records. Someone has to design the systems that capture those signals without asking busy lawyers to fill in forms. The job is part knowledge management, part data engineering and part practice expertise.
Engineers who build on the firm's own data. With agent-based development, a small team can build a working application in days. The engineers a firm needs can take a practice group's problem, build against the firm's matters, clients and documents, and ship something lawyers actually use. They look more like product engineers than traditional application support.
People who govern agents. Agents that read documents, draft and propose changes need permissions, review and an audit trail. Someone has to decide what each agent may touch, review what it proposes, and be able to show a general counsel or a client exactly what happened on a matter. That is a governance role, and it belongs close to risk and to IT.
Product owners for client-facing tools. More firms are building AI tools that clients use directly. Someone has to own each of those tools the way a software company owns a product: what it does, who can use it, how it is supported, and what the client pays for it.
Where IT fits
The friction I see most often is over who gets to build software, now that lawyers and innovation teams can build it themselves with agents. IT departments are built on governance and compliance, and they are right to be. They are also being asked to support a kind of development their standards were never written for: business-side teams building agent-based tools outside the traditional application development process.
The firms making progress treat IT as the owner of that environment. IT defines a governed, separate space where business-side teams can build with agents. It sets which systems agents can reach, what data they can read, and how changes are reviewed before anything touches production. Inside those rules, the business moves quickly. IT keeps the control it is responsible for, and it gains a seat in every AI project the firm runs. I've written more about that partnership in how IT and the business can build a firm's agentic strategy together.
This is also how IT moves from a cost center to a technology-first function. For decades, firms measured value by the lawyer's hour, and IT was overhead that supported the hour. As AI takes on more of the work, value shifts toward how much of what the firm's lawyers know can be captured and delivered to clients. The people who build and run the systems that do that capturing are producing the firm's product. A CIO who owns that platform owns something close to the firm's future revenue.
That shift will take time. Most firms, including the most ambitious ones, still budget IT as a cost. The firms that change that conversation first will find hiring easier, because the strongest candidates want to build the product, and they can tell which firms are offering that work.
What to do in the next two quarters
- Name one owner for AI talent. Innovation, knowledge management and IT often hire separately for overlapping roles. Put one person in charge of the plan so the firm isn't bidding against itself.
- Hire one senior builder before three junior ones. A senior engineer who can direct agents and set the patterns everyone else follows will do more in a year than a team hired for work agents already handle.
- Give IT the charter for business-side building. Write down the environment, the review steps and the permissions, and make IT the owner. It turns a point of friction into a shared program.
- Look inside the firm first. Knowledge management attorneys, practice support staff and paralegals already know where the firm's knowledge lives. Several of the best knowledge engineers I've worked with came from those roles.
- Measure captured knowledge. Track how many matters, positions and precedents are recorded in a form the firm can find and reuse. It is the clearest evidence that the team is producing something the firm keeps.
How Reframe works with these teams
Reframe pairs its platform with senior engineers who work inside a firm's matters alongside the firm's IT and knowledge teams, then hand the work to the firm's own people. Agents build and run applications on the firm's own data, inside the firm's own cloud. Every action they take passes the same permission checks IT controls, and every finding arrives as a proposal a person approves. The goal is a firm that can run the system without us.
AI budgets are rising at almost every large firm. The firms that pull ahead will be the ones that hire and organize the people who turn that spending into knowledge the firm keeps.
Drafted with Reframe Editorial Agent, an AI employee at Reframe. 