A friend asked me a sharp question on a call last week. He wanted to know what actually breaks when you put engineers inside a law firm to build with AI. Is it data access? Culture? Getting authorization right across every system? Or is it the bigger problem, that firms keep bolting AI onto the old way of working when they should be redesigning the work?
My answer surprised him. The firms we engage with at Reframe are past the awareness stage. They know they need to build teams, and budget isn't what's holding them back. Some have told their technology leaders to spend what they need and get in front. The hardest problem is new for everyone involved: letting lawyers and legal engineers build software themselves, under the governance clients expect. Solving it takes IT and the business working together.
IT wants an agentic strategy as much as anyone
Every CIO I talk to wants an agentic strategy for their firm. They see what agents can do, and they want the firm in front. They also run some of the most demanding technology environments in any industry. Client confidentiality, ethical walls, regulatory obligations and outside counsel guidelines all land on their desk, and the governance disciplines their teams have built are a large part of why clients trust firms with their most sensitive matters.
Agentic development brings those teams a new kind of request. A practice group can now describe a tool in plain language and have a working application the same afternoon. A knowledge lawyer can build the intake workflow they've wanted for years. Every one of those builds raises the questions a good CIO should ask. Where does the agent run? Which systems can it reach? Whose permissions does it carry? Who approves the change before it touches client data?
Those are the right questions, and the patterns for answering them are new for the whole industry. The best results I've seen come when IT and the business answer them together. The practice side brings the use cases and the urgency. IT brings the architecture, the controls and the operating discipline that make it safe to scale.
A secure architecture IT and the business build together
We start by bringing IT an architecture early, so both sides design from something concrete. Business-side development happens in an isolated environment, quarantined from production. That environment can reach an explicitly approved list of systems and nothing else. It uses the identities and permissions the firm already manages, so an agent sees only what the person directing it is allowed to see. Nothing moves into production until it passes review and evaluation gates that IT owns.
That design gives everyone what they need. Lawyers and legal engineers build at the speed the tools now allow. IT owns the boundary, the identities, the audit trail and the decision to promote. The firm ends up with an agentic strategy IT can stand behind. IT leaders move quickly once they have a design that meets their standards, and they often become its strongest champions.
A longer shift sits behind the governance question. Most large firms, including the ones leading on AI, still fund technology as a cost center in service of the fee earners. That's starting to change, and the firms that make technology a strategic function will lead.
A longer shift sits behind the governance question. Every large firm, including the ones leading on AI, still runs technology as a cost center that works for the fee earners. Unwinding that will take a while.
When it happens, the economics change. A lawyer's hour matters less than how much of what that lawyer knows the firm can capture and deliver to clients quickly and reliably. The value moves from the time spent to the substance behind it.
Today the most important work is capturing the tacit knowledge of the firm's people: how a senior associate decides which precedent to trust, why a partner accepts a clause in one deal and fights it in the next, which client preferences never made it into a file. Every firm already knows more than any one person in it. That knowledge sits scattered across a dozen systems, and it walks out the door when a matter closes.
Capturing tacit knowledge needs a legal ontology
You can't capture knowledge without a structure to capture it into. A legal ontology is that structure. It's the firm's own model of its clients, matters, parties, documents and provisions, and of how they relate to each other. When the signals coming off the firm's people land in a model the firm defines and governs, the next team can ask about them and trust the answer. When those signals land in a chat log or a shared folder, they're lost.
That's the work we do at Reframe. We build the ontology with the firm's own lawyers and knowledge teams first, then put it to work. Agents build and run applications on it, every action passes the same approval gate, ethical walls bind them, and every finding arrives as a proposal a person approves. The governance architecture and the ontology come from the same principle: the firm decides what exists, who can see it and what can change it.
What large firms are hiring for
The largest firms are trying to hire as fast as they can, and most can't find the people they're looking for. They want people who can design systems that capture the signals coming off lawyers as they work, and who can lead the firm toward using what those systems learn. If you're a technologist considering a move into legal, build that skill set and lead with it in the interview.
Look for a firm whose leadership will fund the build and put technology in front. Not many will, because it's expensive. The ones that do are going to pull away from the rest of the market.
If your firm is working through its agentic strategy, talk with us. We'll bring the architecture to the first meeting.