Developing the AI agent for a world-leading legal tech company

Our client is a rapidly expanding world-leading legal tech company building a collaborative AI workspace for global law firms. The product helps lawyers move faster, think sharper and focus on the work that matters most by reducing friction and automating repetition. With strong momentum and a growing pipeline they needed to strengthen their team with external expertise, and that's where qte came in. Since our engagement started we have helped develop the core of the AI agent – with AI tools as a natural part of that process, of course.

The product is a collaborative AI workspace for legal professionals. It works as a familiar AI chat interface, similar to ChatGPT or Claude, but with deep access to external legal sources connected through APIs and MCPs. This lets lawyers move fluidly between asking questions, pulling in authoritative material and acting on what they find, all in one place.

For the AI agent to be genuinely useful, it needs reliable access to the right external sources. Each new integration is a piece of engineering in its own right: the agent has to know when to reach for a tool, how to call it, and what to do with the results. On top of that, users need a clean and secure way to connect their own accounts. Building this out at the pace the product is growing requires both depth and independence, and with many sources still to be connected there simply isn't enough internal capacity to develop at the desired speed. That's where external support comes in.

We support our client with building the integrations that let the AI agent tap into external legal sources. The work spans the full stack of an integration: designing the agent tool surface so the model knows when and how to use it, shaping inputs and outputs for reliable tool use, and tuning the reasoning loop so the AI can act on the results and help legal professionals reach answers faster. We also build the OAuth and authentication flows that let users securely connect their accounts, so each integration plugs in cleanly without friction.

Beyond the code, the engagement also involves leading technical conversations with external development teams, aligning on scope and unblocking the details that turn a raw API or MCP into a seamless in-product experience.

Building an AI product also means working with AI every day. Since our engagement started, we have worked with multiple AI agents run in parallel throughout the workday, supervised by the consultant in charge leading them with deep domain expertise. What else would you expect from a leading AI product company?