Every Monday at 7am: hours per workday per person, gap days, billable share, entry quality, unbilled time, with time off from the payroll system so a day off isn't flagged as missing time. Reads only, nothing written back.
[ KEY TAKEAWAYS ]
A weekly timekeeping report card from Clio takes five reads and no writes: active users, each person's time entries for the week (date, hours, billable flag, matter, description, created date), expenses for the same week, and, from the payroll system, approved time off. From those, Claude computes five numbers per person: hours per workday, gap days (workdays with no entry, minus time off), billable share, entry-quality flags, and unbilled billable time. It writes one dashboard for the firm and one card per person and changes nothing in Clio. Because it only reads, it can run as a scheduled task on Monday at 7am once the connector runs on a server rather than on someone's laptop.
metrics per person per week. Fewer and the card says nothing; more and nobody reads it.
reads: users, time entries, expenses, approved time off. One tool call each, per person, per week.
writes. Nothing goes back to Clio or to payroll, which is what lets it run with nobody watching.
Monday. The dashboard and the cards are waiting before the first person sits down.
This is the skill a managing partner described to us this week, built by hand in Claude, run by hand because the connector lives on a laptop, and missing one join that made it unfair: paid time off lived in a different system, so every day off read as a missing day. Here is the recipe with that join in, the metrics defined so they can be argued with, and the prompt.
"It goes and pulls everybody's time. Are there gaps? Is the quality of the entries good? What percentage is billable versus non-billable? It comes up with a dashboard, and then report cards for each user. We're trying to make it look at everybody's paid time off in a different system. That part's been a little harder." A lawyer, describing the skill he built by hand, on a call this week.
| Metric | Definition | Source fields | What it catches |
|---|---|---|---|
| Hours per workday | Total hours in the week divided by workdays minus approved time-off days. | Time entry hours and dates; payroll time-off requests. | The person who logged 31 hours across three days and nothing on the other two. |
| Gap days | Workdays with no time entry at all, minus days with approved time off. | Entry dates; time-off dates. | Days that will be reconstructed from memory at month end, badly. |
| Billable share | Billable hours divided by all hours logged. | Billable flag on each entry. | Drift toward admin time; also a staffing signal, not only a discipline one. |
| Entry quality | Three flags: description under five words; hours in whole or half numbers on more than 80 percent of entries; entry created more than two days after its date. | Description, hours, date and created date. | Entries a client would query, and entries written long after the work. |
| Unbilled billable | Billable hours in the week not yet on a bill. | Billable and billed flags. | Work that is done and not yet invoiced, by person. |
Change the thresholds if yours differ. Five words, 80 percent and two days are the ones we would start with; the point is that they are written down, so a card never grades someone on a rule they haven't read.
Seven steps, all reads
Clio's built-in reports give hours and billable totals per user. Use them to check Claude's arithmetic for the first few weeks before anyone acts on a card.
Without it, the first long weekend produces a row of gap days for everyone and the report loses its credibility in a week. Gusto's official connector exposes a list of time-off requests, filterable by status, date range and employee, and the balances per policy. Pull the approved requests that overlap the week, convert them to days, subtract from each person's workday count, and show the subtraction on the card so the person can see it was done. If your payroll system has no connector, a shared spreadsheet of approved days works for the join and is honest about being manual.
The matching is by name, and names drift between systems. The recipe says so on the dashboard instead of guessing, which is the rule for every join in a report people are graded on.
Nothing is written back, and the card goes to the person first This recipe reads Clio and reads payroll. It never creates, edits or deletes anything. That is what lets it run unattended, and it is also the line between a report and a monitoring tool. Send each person their own card before the partner opens the dashboard, and define the five metrics in writing so nobody is graded on a rule they have not seen.
Claude's scheduled tasks run on Anthropic's side, so they can reach a connector that lives on a server and cannot reach one installed on a laptop. This recipe passes the three-question test for scheduled skills: everything it reads is reachable from the cloud once the Clio connector runs on a server and the payroll connector is the vendor's own; it only reads; and no step needs a human answer. Read-only tools do not trip the approval gate that stalls unattended runs with writes in them. The lawyer who built this runs it by hand today for exactly one reason: his connector is on his laptop.
Want this running next Monday?
We put the Clio connector on a server for the firm, point your report skill at it and schedule it with you. The open-source connector is free; the server deployment is a fixed scope. A 30-minute call tells you which shape fits.
Paste this into a Claude skill and adjust the thresholds. It assumes a Clio connector with users, time entries and expenses, and a payroll connector with time-off requests.
Every Monday at 7am, for the previous Monday to Sunday:
1. List active users from Clio.
2. For each user, list time entries in that week: date, hours, rate, total,
billable flag, matter, description, created date. List expenses in the
same week: date, amount, category, matter, user.
3. From the payroll connector, list approved time-off requests overlapping
the week. Match to Clio users by name; if a name does not match, say so
in the dashboard instead of guessing.
4. For each person compute: hours per workday; gap days = workdays with no
entry minus approved time-off days; billable share = billable hours
divided by all hours; entry quality = count of entries with a description
under five words, entries with hours in whole or half numbers only when
that is more than 80 percent of entries, entries created more than two
days after their date; unbilled = billable hours not yet billed.
5. Write one dashboard table for the firm (one row per person, five
metrics, expenses total) and one card per person: the five numbers, the
flagged entries with their dates, and one sentence of context. Do not
write anything to Clio or to payroll. Do not include matter names or
client names in the cards; use matter numbers.What this is not: a reason to start grading people on Tuesday. The first month, run it, hand each person their card, and fix the definitions together. The month after that, it is the report the firm always meant to run and never had time for.
Yes, through a connector that exposes Clio's users, time entries and expenses to Claude. The connector returns each entry's date, hours, rate, total, billable flag, matter and description, per user and per date range. Claude computes the metrics and writes the report. The open-source connector we maintain does this; others expose similar reads.
Join the payroll system's time-off requests for the same week. Gusto's official connector exposes list_time_off_requests, filterable by status, date range and employee, and get_time_off_balances. Subtract approved time-off days from the workday count before computing gaps. Without that join the report is unfair on the first long weekend.
Clio's built-in reports show hours and billable totals per user over a period, which covers two of the five metrics. What they do not do is the gap check against time off, the entry-quality read (short descriptions, round-number clustering, entries logged days late) or the plain-English card per person. Use Clio's totals to check Claude's arithmetic the first few weeks.
Yes, if the connector to Clio runs on a server rather than a laptop. Claude's scheduled tasks run remotely and can reach a remote connector; they cannot reach one installed on a desktop. The skill is read-only, so it does not stall on write approvals. Set it to weekly, Monday, 7am.
It is what every law firm already does by hand at month end, moved to Monday and made consistent. Two rules keep it fair: each person receives their own card before the partner reads the dashboard, and the metrics are defined in writing so nobody is graded on a rule they have not seen. The report never contains matter content, only hours, dates, flags and the person's own descriptions.
The same recipe works with any case system whose connector exposes time entries per user with dates and billable flags. The metric definitions do not change. Our MyCase connector exposes time entries; the expenses and time-off joins depend on what the connector and the payroll system expose.
NEWSLETTER
[ WRITTEN BY ]
Petar Jovanović
Co-Founder & Technical Lead
Co-Founder and Technical Lead at Oktopeak. Builds regulated software for legal and healthcare teams, and leads the rescues of codebases other vendors left half-finished.
[ LEGAL TECH ]
LEGAL TECH
Oct 7, 2026 · 7 min read
Claude already lists your day's work from calendar, email and files. The last step: posting time entries to Clio, the no-duplicate rule, and matter matching.
Read ArticleLEGAL TECH
Oct 9, 2026 · 8 min read
A lawyer sorted his own skills in one sentence: anything with a yes/no gate needs a person, anything that only reads can run alone. The three-question test, his skills sorted, and what the server actually changes.
Read ArticleLEGAL TECH
Oct 7, 2026 · 9 min read
No app can read an iPhone's calls or texts. What does log them to Clio: Clio's own app and texting, Quo, Call Connector, VXT, Corvum, LegalMate, Time Miner.
Read Article[ GET STARTED ]
30-minute call. No pitch deck. Just an honest conversation about your project.