AI Meeting Notes for Lawyers: Otter vs Fireflies vs Custom

By Jude Lee · · Comparison

Attorneys and a paralegal in a client consultation with a laptop recording the meeting

The part of a client meeting that actually costs you time

Recording and transcribing a consultation is a solved problem. The expensive part happens afterwards: someone re-reads the transcript, writes a matter note in the firm’s house style, flags the limitation-period question, opens a task for the conflict check, drafts a follow-up email, and captures the time. That post-meeting tail is the workflow — not the transcript.

That distinction maps onto the difference between a generative tool and an agentic one. A summarizer produces text you then act on: the output is a document, and every downstream action is still yours. An agent takes actions in your systems — it writes the note into the matter, creates the task, populates the draft — which is why it needs credentials, permissions and supervision that a summarizer never does. Most of the disagreement about “AI for meetings” is really a disagreement about which of those two things you are buying.

Route one: general-purpose notetakers

Otter, Fireflies, Zoom’s built-in AI companion and Microsoft Teams recap all do the core job: join the call, transcribe, produce a summary and action items. They are cheap per seat, deploy in an afternoon, and work across every meeting type — client consults, expert calls, internal case strategy.

What they don’t do is understand your matter. The summary lands in the notetaker’s own workspace, not in Clio, MyCase, Smokeball or NetDocuments. Someone still copies it across. And because they’re horizontal products, the summary format is generic: “key points, action items” rather than “facts, damages, jurisdiction, statute-of-limitations risk, conflicts to run.”

Before you deploy one firm-wide, check the vendor’s current documentation on data retention, model training and where recordings are stored — and whether your plan tier changes those answers. This is also the category where individual staff quietly sign up with personal accounts, which puts client conversations on a consumer plan nobody in the firm has reviewed.

Route two: whatever your practice-management platform already ships

The legal platforms are adding AI summarization and note features of their own — Clio Duo and Smokeball’s Archie are the two most visible examples for small firms, with MyCase and other suites shipping comparable assistants. The built-in option has one enormous advantage: the data never leaves the system where your matters live, so filing the note to the right matter is trivial rather than an integration project.

The trade-off is scope and pace. Built-in features do what the vendor decided to build, in the vendor’s format, on the vendor’s timeline — and they generally cover meetings that happen inside that platform’s orbit. These feature sets change quarterly, so treat any capability list you read (including this one, written as of early 2026) as stale and verify the current scope against the vendor’s own documentation before you buy. We compared the platform-native assistants in more detail in Clio Duo vs Smokeball Archie vs a custom AI agent.

If your only complaint about the built-in feature is the summary format, don’t build anything — write a better prompt template and move on.

Route three: a custom agent wired to your systems

Here the pattern is: a transcript (from whatever source you already trust) plus an AI assistant such as Claude, plus MCP — the Model Context Protocol, an open standard for giving an AI governed access to specific tools and data. Through an MCP connection to your practice-management system, the assistant can look up the matter, write the note, create tasks and draft the follow-up. The mechanics of that connection are covered step by step in connecting Claude to Clio with MCP.

The second half of a custom build is the skill — a packaged, reusable instruction set that teaches the assistant to produce your firm’s consult note the same way every time: your section order, your intake fields, your rule that any date mentioned gets flagged for calendaring rather than calculated.

Be honest about what that costs. Scoping, building and testing a single workflow is measured in weeks of someone’s attention, not an afternoon, and the bill doesn’t stop at launch: when your practice-management vendor changes an API field, deprecates an endpoint or ships its own competing feature, someone has to update the MCP server and re-test the skill. Ask before you start who that someone is, whether the work is documented well enough that a second person could pick it up, and what happens if the person who built it leaves or the outside developer moves on. A custom agent nobody can maintain is a liability wearing a productivity costume.

Off-the-shelf notetaker
Live in a day. Low per-seat cost. Works for every meeting type. Generic output format. Summary lives outside your matter file. No actions taken in your systems. Vendor controls retention and roadmap.
Custom agent + MCP
Weeks of build and testing, plus a named owner for ongoing maintenance as APIs change. Output shaped to your practice area. Writes directly to the matter, opens tasks, drafts follow-ups. You choose the hosting and set the access rules — and carry the responsibility when it misfires.

Where these agents break

Be blunt with yourself about failure modes before you scale one:

Deciding whether the build pays

Don’t take anyone’s headline savings number, including ours. Model it with numbers you can actually measure this month:

  1. Measure the tail, not the meeting

    Time three people writing up a consult end to end — note, tasks, follow-up email, time entry. Record the real number in hours, not an estimate.

  2. Multiply by volume

    Consults per month × hours per write-up = monthly hours at stake. Split it by who does the work; paralegal hours and attorney hours carry very different costs.

  3. Apply your own rate

    Use your firm’s blended billable rate for attorney time and loaded cost for staff time. Rates span an enormous range by practice area and market, so use your figures, not a national average you read somewhere.

  4. Discount for review

    Assume the agent gets you a strong draft, not a finished product. Cut the modelled savings by whatever share of time review still takes — a 30–50% haircut is a conservative planning assumption, not a measured finding.

  5. Add the capture upside — carefully

    Faster follow-up may improve conversion and reduce unbilled time. Model it as a sensitivity range you can test, not a promised gain.

In short: model it as hours × your rate, discount 30–50% for review, and scope the first build to consults only.

A fair read: if you run a handful of consults a week and your write-ups are already tidy, a low-cost per-seat notetaker plus a good template is the right answer, and building anything is a waste of money. If you run consults at volume, in a repeatable practice area, and the write-up backlog is genuinely costing you conversions or billable time, the custom route starts to make sense — and even then, start with the transcript source you already have and build only the filing-and-drafting layer.

Pick the job before you pick the tool

There is no single “best AI program for a law firm,” and asking for one is usually a symptom of shopping before scoping. Harvey and CoCounsel target research and analysis work; Otter and Fireflies target meetings; the practice-management assistants target the data already inside their platform; Claude, ChatGPT and Copilot are general assistants that become firm-specific only when you connect them to your systems and define skills. Name the job first — new-client consults, discovery triage, billing narratives — then pick the narrowest tool that does that job under your firm’s confidentiality and supervision rules. Firms that shop for the best all-purpose legal AI tend to end up with three overlapping subscriptions; firms that ask what their consult write-up actually costs them tend to fix something.

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