AI Client Update Agents: Clio vs MyCase vs Custom

By Jude Lee · · Comparison

Paralegal and attorney reviewing a drafted client status update email on a monitor in a law firm office

The work that leaks first

Ask any small-firm managing partner which task gets dropped when a trial date moves, and it’s the same answer: proactive client communication. Nobody bills for “just checking in,” so it waits. Then the client calls, the paralegal stops what they’re doing, and the lawyer reconstructs a matter they last touched three weeks ago — call it ten or fifteen minutes, but that figure is purely illustrative. Time yourself once and use your own number; everything below depends on it.

This matters beyond annoyance. ABA Model Rule 1.4 imposes an affirmative duty to keep clients reasonably informed about the status of a matter and to promptly comply with reasonable requests for information — your state’s version controls, so check your own rules. Communication failures are also a well-known source of client grievances. Whatever the exact picture in your jurisdiction, the operational point stands: the cheapest complaint to prevent is the one caused by silence.

So this is a good automation candidate — not because AI writes beautiful prose, but because the bottleneck is assembly, not writing. Someone has to look at five systems and figure out what changed.

What a client-update agent actually does

A chatbot answers a question you type. An agent carries out a multi-step job: it queries your systems, assembles facts, drafts an artifact, and routes it somewhere. For client updates, the loop looks like this:

  1. Pull the week's activity

    Query the matter record for new documents, filed pleadings, calendar changes, completed tasks, payments received, and incoming correspondence since the last update.

  2. Decide what's client-relevant

    Most activity isn’t. An internal research memo usually isn’t update-worthy; a continued hearing date always is. This filtering logic is where the real value — and the real build work — lives.

  3. Draft in the firm's voice

    Plain English, no jargon, a clear “what happens next” line, and an explicit “what we need from you” if anything is outstanding.

  4. Route for human approval

    The responsible attorney reviews and edits before anything reaches the client. No exceptions on this one.

  5. Send and log

    Deliver via portal or email, then write the communication back to the matter file so it’s part of the record.

Option 1: what you already pay for

Clio and MyCase are the default answers for case management software for small law firms, and both now ship AI features — Clio Duo and MyCase IQ respectively — alongside client portals, secure messaging, and text. Treat the specific capabilities as a moving target and verify against each vendor’s current product documentation and release notes before you buy; AI feature sets in practice management change frequently.

The honest strengths: zero integration work, the data is already there, the ethics surface is smaller (one vendor, one DPA, one security review), and the per-user cost is bundled. If your updates are mostly “here’s what’s on the calendar and here’s your invoice,” rule-based workflows plus a good template may get you most of the benefit with no AI at all. That’s a legitimate outcome, and often the right one.

The limits: built-in AI sees what’s inside that platform. If your documents live in NetDocuments or SharePoint, your client email lives in Outlook, and your medical records sit in a third system, the platform’s assistant has a partial view — which means a human still has to assemble the rest. Our broader comparison of Clio Duo, Smokeball Archie, and a custom agent goes deeper on that ceiling.

Option 2: a custom agent over MCP

MCP — the Model Context Protocol — is an open standard for giving an AI assistant governed access to your tools and data. Practically, it means an assistant like Claude can be granted read access to your practice management system, document store, and calendar through connectors you control, with scoped permissions and an audit trail. We walk through the mechanics in connecting Claude to Clio with MCP.

For client updates, the custom route buys you three things the built-ins generally can’t: cross-system visibility, firm-specific filtering rules (“never mention settlement figures in a written update without partner sign-off”), and a reusable skill — a packaged set of instructions that makes the agent produce the same structure every time, for every practice area, no matter who runs it.

The cost side is real: someone has to build and maintain the connectors, write and test the filtering logic, and own it when the API changes. That’s a project, not a toggle.

Built-in platform AI
Fast to switch on. No integration work. Single vendor for security review. Limited to data inside the platform. Behavior changes when the vendor updates the feature. Best when your workflow already lives almost entirely in one system.
Custom agent via MCP
Sees across practice management, documents, email, calendar. Encodes your firm’s own rules about what clients hear and when. Every connector you add widens the confidentiality footprint — more systems reachable from one place — so budget for its own security and access review alongside the build. Requires an owner and ongoing maintenance. Worth it when cross-system assembly is the actual bottleneck.

Start narrow: one matter type, not the whole book

The old lawyer’s heuristic — that a small share of clients and matters drive most of the revenue — is folk wisdom, not a measured law, and I’d treat it that way. But it’s a useful filter here. Don’t automate updates for every matter at once. Pick the matter type where updates are most repetitive and most frequently skipped (volume PI, immigration, estate administration, collections defense), build it there, and leave bespoke litigation to the attorney’s own pen.

If the update requires judgment about strategy, a human writes it. If it requires assembling facts that already exist in your systems, an agent should draft it.

The economics, modeled honestly

I have no benchmark hours for your firm, so build the estimate yourself. The formula:

(minutes per update ÷ 60) × updates per month × fully-loaded hourly cost = current cost of doing it manually.

Then subtract the time still required for review — realistically a few minutes per message, not zero — plus your monthly tooling or amortized build cost. The third term is the one most firms forget: recovered capacity only turns into money if it’s redeployed into billable or business-development work, not absorbed into the day.

min × count × rate
Model your own current cost before comparing options
Review time ≠ 0
Attorney approval is a permanent line item, not a phase

There’s also a revenue side I can’t quantify for you — so treat it as a hypothesis to test rather than a benefit to assume. Before you roll out, record inbound “any update?” call volume and average days-to-payment for one matter type; re-measure after a quarter of predictable monthly notes and see whether either moves. If they don’t, you’ve still bought back assembly time. If they do, you have a number that’s actually yours.

What firms are actually running

The realistic 2026 stack at a small or mid-size firm is: a practice management platform with AI features switched on, a general assistant (Claude, ChatGPT, or Copilot) for drafting, maybe one specialist tool for research or contract review, and — at the more sophisticated end — one or two custom agents wired into firm systems. That’s it. Anyone describing a wholesale AI transformation of a twelve-lawyer firm is selling something.

The question worth asking isn’t which tools you own; it’s whether partner hours are shifting from administrative assembly toward matters and origination.

Before you build anything

Read ABA Formal Opinion 512 (2024) on generative AI tools, which addresses competence, confidentiality, client communication about AI use, and fees — then check your own state bar’s guidance, which may be stricter. Decide who approves outbound messages, keep that oversight documented, and start with one matter type. For a fuller framework covering both who signs off on what and how to model the cost of that review, see our breakdown of AI agent oversight models.

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