AI Motion Drafting: Word Copilot vs CoCounsel vs Custom

By Jude Lee · · Comparison

Two lawyers reviewing a printed motion draft next to a laptop in a law firm conference room

What “AI motion drafting” actually means now

Strip the marketing and there are three separable jobs inside drafting a motion:

  1. Structure — caption, standard of review, headings, procedural history, relief requested. Highly formulaic, jurisdiction-specific, and the easiest thing to automate well.
  2. Argument assembly — pulling the fact section from the record, matching the argument to authority, echoing language that has worked for your firm before. Partially automatable, heavily supervised.
  3. Authority verification — does the case exist, is it still good law, does it say what the draft claims. Not safely automatable end to end today.

Most tools are sold as if they do all three. The honest framing: AI can get you a competent first draft of (1) and a rough (2); (3) needs a tool that checks against a real citator plus a human who reads the cases.

Option 1: Word with Copilot — the default that’s already paid for

If your firm runs Microsoft 365, you may already have a drafting assistant sitting in Word. It can restructure a section, tighten prose, summarize a long declaration, and generate boilerplate. It may also be able to reference documents held in SharePoint or OneDrive — but that depends on your license tier and tenant configuration, so verify against Microsoft’s current documentation rather than a sales deck.

What it does not do: know Ninth Circuit summary judgment standards, check whether a case is still good law, or reliably follow your local rules on citation format. It’s a writing assistant, not a legal research system. For a solo doing a routine motion to compel with a template they already trust, that may be entirely sufficient — and “sufficient and already paid for” beats “excellent and quoted per seat” more often than vendors admit.

To actually get value from it:

CoCounsel (Thomson Reuters), Harvey, and similar products are built on legal corpora and connect to research databases. That connection is the real product: a draft that cites to a live citator is categorically different from a draft that cites to a language model’s memory. They also tend to ship workflow scaffolding — document comparison, deposition review, research memos — so drafting is one tile in a broader suite.

The trade-offs are the usual ones for per-seat legal software: you adapt to the vendor’s workflow, output is calibrated to a general legal style rather than your firm’s, and pricing is typically quoted rather than published. We compared the suites in more depth in Harvey vs CoCounsel vs a custom AI agent.

If you pilot one:

Option 3: A custom agent over your own brief bank

Your best motions are already written. They’re sitting in iManage, NetDocuments, SharePoint, or a folder called MOTIONS - GOOD ONES. A custom drafting agent is an AI assistant (Claude, for example) connected to that repository through MCP — the Model Context Protocol, an open standard for giving an AI governed, permissioned access to your systems and tools — plus a skill: a packaged, reusable instruction set that tells the assistant exactly how your firm drafts a motion.

The skill encodes things a general tool can’t know: that your opposition briefs lead with the procedural posture, that a particular judge wants a one-page summary of relief up front, that you never cite unpublished state decisions, that fact sections cite to Bates numbers in a specific format.

A drafting agent inherits the quality of the briefs you feed it. Curation is the actual work — and it’s partner-level judgment, not an IT task.

This is the same architecture described in building a custom MCP server over your matter data — read-access to the right documents, scoped by matter and ethical wall, with an audit trail. Whether your DMS supports this cleanly matters a lot; see iManage vs NetDocuments vs SharePoint for AI agents.

Off-the-shelf drafting features
Live in days. Vendor handles updates and security posture. Research-database grounding bundled in, without you negotiating separate data licensing. Per-seat pricing scales linearly with headcount. Output is generically legal — you’ll still edit voice and structure. Best when your motion practice is varied and you file across many jurisdictions.
Custom drafting agent + MCP
Weeks of setup plus ongoing ownership. Drafts from your filed work product, so the first pass is closer to final. You control exactly which documents it can read. No per-seat ceiling. You still need a citator subscription and a verification step. Best when you file the same motion types repeatedly and have a deep brief bank.

Building the custom path without breaking anything

  1. Pick one motion type you file constantly

    Motions to compel, opposition to summary judgment, motions in limine — whichever you’ve filed at least a dozen times. One type, one jurisdiction.

  2. Curate a gold-standard set

    Have a partner select the 8–15 best filed examples. Bad briefs in, bad briefs out.

  3. Write the skill

    Document how your firm drafts this motion: section order, tone, citation conventions, what never appears. Plain language, not code — a senior associate can write it in an afternoon.

  4. Connect the data, scoped

    Use MCP to give the assistant read access to the matter file and the curated brief set — not the whole DMS. Scope by matter, respect ethical walls, log every access.

  5. Route every citation through verification

    The draft goes to a citator check and a human read. No exceptions, regardless of vendor. See our breakdown of AI citation checking options.

  6. Measure against your own baseline

    Time three drafts the old way, three the new way, same attorney. Your numbers, not a vendor’s case study.

What actually goes wrong

Skepticism about AI legal workflows is warranted, and here is my honest list of the disadvantages:

Automate the scaffolding, keep the argument human

Take a motion to compel. The caption, the meet-and-confer certification, the discovery-history recitation, the Rule 26 proportionality standard, the relief requested, and the proposed order are substantially reusable across every one you’ve filed. What isn’t reusable is the paragraph explaining why these specific objections fail against this record. That’s the split worth designing around: automate the recurring scaffolding aggressively, and spend the recovered attention on the two or three paragraphs that decide the motion.

Sizing the payoff with your own numbers

Don’t take anyone’s hour-savings claim, including ours. Model it:

hours saved × billable rate
Recovered capacity per motion — use your own timekeeping data
motions per month × savings
Monthly effect, before tool cost
minus seats × price (or build + maintenance)
Net — run it for both options

The more interesting question isn’t cost savings — it’s what the recovered hours get reallocated to. Drafting time converted into more matters handled, faster turnaround on fixed-fee work, or partner time spent on origination is worth more than the same hours saved and left idle. Our automation ROI framework walks through the full calculation.

There’s also a legitimate floor case. If you file a handful of motions a quarter, or you’re writing an appellate brief on an issue of first impression, none of these three options earns its setup cost or its risk — the work is bespoke reasoning, and the scaffolding you’d automate is a small share of the effort. Good templates and a careful human are the right answer.

As of early 2026, pricing across legal AI suites is still moving and MCP support varies by document management vendor. Re-verify capabilities directly with product documentation before committing to a multi-year seat contract.

Where is your firm losing billable hours?

Get a free automation audit: we map your intake-to-invoice workflow and show you exactly what's worth automating — before you spend a dollar.

Get a free automation audit

Related guides