Automatic Time Tracking for Lawyers: Clio + Outlook Test

Most lawyers don't lose billable time in big chunks. They lose it in the 20-minute call that never made it into Clio and the six client emails sent between hearings. Clio's own benchmark data, as cited on Clio's blog, puts average utilization at about 38%, or roughly 3 billable hours in an 8-hour day. Clio's 2026 Legal Trends Report, as summarized by LawSites, projects 42% for 2026, up from 39% in 2025. Either way, that's a vendor reading its own data, and it doesn't say how much of the gap is unrecorded work versus admin. Still, anyone who has rebuilt a Friday's time from a sent folder knows some of it is.

So we built a small workflow to do that rebuild every evening, and tested it. This is a write-up of what we built, how it behaved in testing, and what it can't do. We tested against mock versions of Clio, Outlook and the email sender, not a live firm's accounts, and we say more about what that means below.

What we built

The daily time capture recipe for Clio runs at 6:30 p.m. on weekdays. It reads the day's Clio calendar and the attorney's sent Outlook mail, tries to match each item to exactly one open matter, and creates unbilled time entries in Clio labeled [Auto-draft - review before billing]. Anything it can't match safely goes into a review email instead.

The workflow has 13 nodes. We don't have screenshots of it yet, so here they are in order, taken from the workflow file:

Weekdays 18:30 (schedule), plus a manual/test run webhook
Day window + calendar query, then Clio: calendar entries
Sent-mail query, then Outlook (Microsoft Graph): sent mail
Matters query, then Clio: open matters
Draft time entries (the rules)
Entries to create, then Clio: create time entry
Review email, then send the review email through Resend

The rules are short enough to state in a paragraph. A calendar event linked to a matter becomes an entry for its actual length, rounded up to the next tenth of an hour. A sent email to a client who has exactly one open matter becomes a 0.1-hour entry. Emails to your own domain are skipped. Everything else goes to the review email: matters that require UTBMS task codes, clients with two open matters, recipients the workflow can't find in Clio.

There's no AI model involved. Every decision is a rule you can read in the workflow.

How we tested it

Our testing notes describe the setup for every recipe. We imported the workflow into n8n 2.45.4 and ran it against a local mock server built from Clio's and Microsoft's API documentation. The mock checks the same request shapes and required fields as the real APIs, enforces Clio's rate limit with HTTP 429 replies, and records every call. The matters and clients are invented. The last run, on October 28, passed 5 of 5 scenarios:

T1, a busy Tuesday. Two meetings and two client emails became draft entries. An email on a matter that requires UTBMS codes and a CLE registration email to someone with no matching open matter were kept out of Clio and listed for review. A lunch on the calendar and a discovery-index item were left out of both Clio and the review email.
T2, running the same day again. No duplicate entries and no second email.
T3, a Monday. A 24-minute call rounded up to 0.4 hours, but its matter needed UTBMS codes, so it went to review rather than into Clio.
T4, Clio rate-limits a call. The run waited, retried and finished. A day with nothing on it sent nothing.
T5, a client with two open matters. The email was not guessed onto either matter. It was listed for review.

The entries the mock received look like what you'd expect: a 30-minute custody-schedule call was posted with quantity: 1800, because Clio's API has measured time-entry quantity in seconds since version 4.0.4. That detail is easy to miss if you work from older examples, and a value of 0.5 would be read as half a second.

What testing against mocks can't tell you: whether your firm's real calendar is tidy enough to match, how many of your emails go to clients with more than one matter, and how the drafts compare with what you'd have billed. The recipe's setup steps suggest running it for one past day with the manual webhook and comparing before you turn on the evening schedule. We'd do the same.

What it doesn't capture

The gaps are real and worth stating plainly:

Phone calls, drafting and research that don't leave a trace in the calendar or sent mail.
Time on each email. Every email is a flat 0.1 hours by default (EMAIL_MINUTES=6), not the time you spent. You can change it, but it's still a flat number.
Clio has no draft state for time entries. The entries are created unbilled with a label, and someone has to review or delete them before billing.
It reads one page of up to 200 events, emails and matters per day. A large firm would need pagination added.

A word on billing ethics

The flat email default is the part to think hardest about. In ABA Formal Opinion 93-379, the ABA said a lawyer "may not bill more time than she actually spends on a matter, except to the extent that she rounds up to minimum time periods (such as one-quarter or one-tenth of an hour)." A tenth of an hour for a one-line email fits that, but a tool that drafts entries doesn't change who is responsible for them. Treat every auto-drafted entry as a suggestion you confirm, follow your engagement letters and your state's version of the rules, and ask your bar's ethics hotline if you're unsure. This is general information, not legal advice.

How this compares with Clio's own options

Before building anything, check what you already pay for. According to Clio's pricing page, Gmail and Outlook email integrations that let you track time from your inbox are on every plan, and so is UTBMS/LEDES billing.

Clio also has a newer feature, Clio Timekeeping, that its help center describes as a pilot. It uses AI to draft time entries from your activity, which can show up about 15 minutes later, and you review them and switch on "Bill-ready" for the ones you keep. The Clio Desktop App extends it to work you do outside Clio on your computer. LawSites' recap of ClioCon 2026 covers what Clio announced around it. We haven't used it, so we can't compare results. If you're eligible for the pilot, try it before building your own.

The case for a workflow like ours is narrower: you want written rules you can read, you want ambiguous items in an email instead of in Clio, or you want to change exactly what counts. One catch on plans: Clio's pricing page lists "Apps and integrations" as not included on Starter, so this recipe, like any API workflow, needs Core or above.

What it costs

There's no AI model and no new subscription. You need Clio Manage on a plan with integrations, an n8n instance (free self-hosted, or a Cloud plan from €20 a month according to n8n's pricing page), and an email sender. At about 20 working days a month, the review email is roughly 20 messages, which fits in Resend's free plan. The recipe page has the details.

One setup point to know about: the workflow refreshes the Clio token itself, and the Clio client ID, secret and refresh token sit in n8n environment variables. Any workflow on the same n8n instance can read those, so run it on an instance only your firm uses.

All law firm automations
Invoice and retainer reminders, and our comparison of Clio Payments and LawPay for trust retainers
Weekly client status updates
Inquiry intake into Clio, with the build walkthrough in Clio intake automation with n8n and the after-hours inquiry auto-reply
Missed-call text-back for law firms and the ethics checklist for texting clients
Browse every recipe

We tested automatic time capture for Clio and Outlook