The Integration I Didn't Have to Build

Allan Rees-Bevan
Managing Director, Morae Australia

I spend a lot of my working life sitting between lawyers and the systems they use every day: document management platforms, practice management systems, drafting tools. For years, the pitch from every legal tech vendor was some version of “we connect everything together.” A firm wanted a client’s matter details to flow into a document, so someone built a bridge. Then they wanted it to flow the other way, so someone built another one. Every upgrade broke something, and every new tool meant another six-month integration project.

That’s finally starting to change, and it’s changed how I work day-to-day. The reason is a deceptively simple piece of plumbing called the Model Context Protocol, or MCP.

What MCP actually is

MCP is an open standard that lets an AI assistant, whether that’s Claude, Copilot, or something else, discover and use external tools directly, without a developer hand-coding a connector between them. Think of it as a USB-C port for AI: instead of a different custom cable for every device, you get one standard socket that anything can plug into.

For legal teams, that matters because two of the tools lawyers already rely on, iManage Work and Smarter Drafter, have both built MCP servers. iManage now offers a governed, permission-bound gateway into your firm’s documents and matter knowledge. Smarter Drafter has done the same for document automation, connecting an AI agent to your firm’s actual approved templates and clause library rather than letting the AI generate legal text from scratch. Once both are switched on, an AI agent can move between them in a single conversation.

Why this made my life easier

Before MCP, if I wanted an AI tool to help me pull together a matter summary, I was copying and pasting between three or four browser tabs: iManage to find the right documents, another system for matter data, a drafting tool for the output, and back again to save it. It worked, but it was tedious, and it was easy to lose track of which version of a document was current.

Now I can ask Claude something like “pull together the key correspondence and the current draft agreement for the Nguyen matter” and it goes and finds them in iManage directly, respecting whatever permissions and information barriers already apply to me. I’m not exporting anything, and I’m not giving the AI a bulk copy of the document library to rummage through. It only sees what I’m already allowed to see, and every retrieval is logged, ensuring transparency and auditability for every AI action taken via MCP. That distinction matters enormously in a profession where confidentiality isn’t optional.

Where it gets genuinely useful for lawyers

The real payoff shows up when iManage Work and Smarter Drafter are combined in the same AI conversation. Here’s a realistic version of how that plays out.

A lawyer needs a cost agreement for a new matter. Instead of opening Smarter Drafter, hunting for the right form, and manually keying in client details, they simply tell their AI agent, “draft a cost agreement for the Patel matter”. The agent looks up the matter in iManage, pulls the client name, matter number, and fee arrangement already on file, and identifies the correct cost agreement template in Smarter Drafter. If a detail is missing, say, the hourly rate hasn’t been recorded, it asks the lawyer directly rather than guessing. Once it has what it needs, Smarter Drafter assembles the document from the firm’s own approved precedent, not from text the AI invented, and the finished agreement comes back into the conversation. From there, the same agent can save it straight back into the correct iManage workspace.

Two prompts. No tab-switching. No field mapping. And critically, no risk of a hallucinated clause slipping into a client-facing document, because the operative text always comes from Smarter Drafter’s deterministic assembly engine, not from the AI’s imagination.

Smarter Drafter designed it this way on purpose - it's not luck. Their view is that AI is probabilistic and legal drafting cannot be. So the AI handles the conversation and the intent, and Smarter Drafter’s deterministic engine produces the operative text from the firm's own approved precedent. AI builds it. Determinism runs it.

I’ve seen a similar pattern work in reverse for research and review. A lawyer preparing a new commercial lease can ask the agent to find comparable precedents already sitting in iManage, use them to sense-check unusual clauses in a draft, and then flag anything that departs from the firm’s standard position, before the document even goes to a partner for review.

Why lawyers should care about this now

The obvious objection to AI-assisted drafting has always been trust. Generative AI is fluent, but fluency isn’t the same as accuracy, and a confidently wrong clause in a client document is a real professional liability risk. What MCP does is separate the two jobs an AI agent needs to do well: understanding what you’re asking for, and producing text you can actually rely on. The conversational intelligence comes from the AI. The precision comes from Smarter Drafter’s firm-approved templates. The governance comes from iManage’s permission and audit layer. None of them are doing the other’s job.

There’s also a quieter benefit for anyone who has ever had to justify legal tech spend to a managing partner. Firms have historically paid, in both time and money, to build and maintain the custom integrations between their PMS, DMS, and drafting tools. MCP replaces that patchwork with one open connection per system. It’s not just faster for lawyers day-to-day, it’s cheaper to maintain for the firm.

Getting started

If your firm already uses iManage Work or Smarter Drafter, the MCP capability is available now. In most cases it is a matter of enabling it through your IT or knowledge management team rather than buying anything new. If you’re using Claude, Copilot, or similar already, ask what’s connected behind the scenes. The most useful AI tool in your firm right now might not be a new subscription. It might just be the one you already have, finally able to talk to the systems you already trust.