How to Run a Weekly Review with AI

A solo lawyer billing 32 hours a week has no spare Friday to reflect. Client calls, court deadlines, and unsigned retainers stack up until Monday arrives and the week is already spent. The review that would have caught the overdue invoice and the stalled matter never happens, so the same three problems resurface every month.

This workflow turns a model into the clerk of your week. You talk in plain notes; it sorts, counts, and asks the questions you skip when you are tired. Twenty minutes later you have a cleared list of open loops, a shortlist of matters that actually need you, a realistic five-day plan, and the follow-up messages you have been dodging.

The Output, Before You Start

A one-page review sheet with four blocks: what moved, what is stuck, what it cost, and what happens next. Short enough to reread in ninety seconds on Monday morning. The goal is not to feel productive. The goal is to walk into next week knowing which matters are quietly bleeding time and which are ready to close.

Step 1: Dump the Week Without Editing

Resist the urge to organise first. Messy input gives the model something to structure; a tidy input usually means you edited away the details that matter.

You are my weekly review partner. I am a solo lawyer running a two-person practice: me plus one paralegal. Today is Friday, and this review covers Monday through Friday.
I will paste a rough dump of the week in no particular order. Do not summarise or restructure yet. When I am finished, ask me up to 4 questions about anything ambiguous, then stop and wait for my answers.
Assume these constraints: Mondays are court mornings, Friday afternoons are blocked for drafting, and my billable target is 30 hours a week.

You get a short interview instead of a wall of output. The questions usually surface what you forgot to log, such as a call that ran long or a matter a third party is still sitting on. Answering four questions now prevents a review built on half the facts.

Step 2: Sort the Open Loops

The value of a review is separating work that is closed from work that only feels closed. Force the model to place every item into one of three buckets.

Take my dump and sort every item into three lists:
1. Closed this week.
2. Open and waiting on someone else.
3. Open and waiting on me.
For each open item add: the client or matter it belongs to, the next physical action, and how many days it has been open.
Flag anything open more than 10 days as at risk. Do not invent client names; use the ones I gave you.

You get a short “waiting on me” list, which is almost always the shorter and more honest one. In a small practice that list tends to hold two or three items that carry real risk, like an unsigned engagement letter or an unfiled motion.

Step 3: Reconcile Hours and Invoices

A weekly review without numbers is a mood, not a measurement. Ask the model to account for time and money using only the data you provide.

From the dump, estimate where my week went and group hours into billable and non-billable.
Then list: (a) matters where more than 30 minutes went unbilled, (b) invoices sent but unpaid past 21 days, and (c) any matter with low activity that a client may perceive as stalled.
Use approximate hours when I gave ranges. Do not add other clients or invent figures.

You get two lists that change behaviour: the billable leaks and the ageing invoices. Seeing that a single matter ate four hours with nothing to invoice is the nudge that fixes next week’s timekeeping.

Step 4: Pick Three Priorities for the Next Week

Ten priorities guarantee none get done. Cap the list at three and make each one an outcome rather than a chore.

Recommend exactly 3 priorities for next week. Each must be an outcome, not a task, and each must map to the "waiting on me" list from earlier.
For every priority give: why it matters, what done looks like, and the first 30-minute action I can take to start it.
Respect that Monday morning is unavailable and Friday afternoon is reserved for drafting.

You get three outcomes with a defined first move, so Monday starts without a planning session. A client-ready draft beats “work on the Smith matter” because it tells you when to stop.

Step 5: Draft the Follow-Ups You Keep Avoiding

The messages you postpone cost more than the work itself. Let the model write the first version so editing is faster than starting.

Draft three short messages using plain language and no legalese:
1. A polite second reminder to a client whose invoice is 34 days overdue.
2. A status update to a client whose matter is waiting on a third party.
3. A handover note to my paralegal listing the three tasks I am assigning this week.
Keep each under 120 words. End the invoice reminder with a specific next step, not a vague request.

You get three drafts you can send after a read-through. The legal skill is in the wording you adjust, not in the blank page you avoided.

Step 6: Freeze It into a Reusable Template

A good review is a ritual, not a one-off. Capture the shape so next Friday is faster.

Convert everything above into a reusable weekly review template with blanks I fill each Friday.
Include the 4 interview questions, the three-column sort, the hours and invoice check, and a slot for the 3 priorities.
Keep the whole template under one page and add a one-line instruction at the top explaining how to use it.

You get a portable template you can paste into any chat next week. Because the structure is fixed, the model only has to process fresh input rather than rebuild the method.

What a Finished Review Contains

• A three-column sort of every item from the week: closed, waiting on others, waiting on you.
• An at-risk flag on anything open more than ten days.
• An hours breakdown split into billable and non-billable.
• A list of unbilled work and invoices overdue past three weeks.
• Exactly three outcomes for next week, each with a first 30-minute action.
• Three drafted follow-ups, under 120 words each.
• A reusable one-page template for next Friday.

Where Weekly Reviews Fall Apart

Pitfall 1 — reviewing a curated version of the week. If you only log the wins, the open loops stay invisible. Keep the dump messy and complete; the sort is the model’s job, not yours.

Pitfall 2 — trusting invented numbers. A model will happily estimate time or fees you never gave it. State that it must use only your figures, and correct any it adds.

Pitfall 3 — turning the review into a to-do list. A list of twenty tasks is the problem, not the fix. Three outcomes with a clear finish line is the whole point of the cap.

Pitfall 4 — skipping the follow-ups. The overdue invoice and the waiting client are exactly the items the review is meant to catch. If you do not send the drafts, the review was entertainment.

If the review keeps surfacing the same priorities, pair it with our weekly work plan prompt to turn outcomes into daily blocks. To clear the inbox that feeds the review, see inbox and calendar prompts, and browse more systems in our productivity prompts list.

A Friday in Practice

Here is what one line of the finished review looks like, so you can see the output is specific rather than generic.

Matter: Hartley v. Nolan (contract dispute)
Status: Open, waiting on me
Next action: sign and send the amended engagement letter
Days open: 14 (at risk)
Hours this week: 4.5 (unbilled)
Note: client emailed twice; no reply logged

Two lines like this at the top of a review do more than any productivity app. They name the client, the exact next action, and the cost of delay in one glance.

Questions That Come Up

How long should a weekly review take? Twenty minutes if the dump happens during the week. If you are reconstructing five days from memory on Friday, the review will take an hour and will still miss things.

Can I run this if I am not a lawyer? Yes. Keep the six steps and swap the reconciliation block for whatever you measure: tickets closed, deals advanced, or study hours completed.

What if I have no billable work to reconcile? Replace Step 3 with a check on deliverables promised versus delivered. The point is an objective count, not money specifically.

Should I keep every review? Keep one paragraph per week in a single file. Reading four consecutive reviews makes patterns obvious that a single review hides.

Sources & Notes

The three-column open-loop sort and the weekly cadence are long-standing habits from personal knowledge management; the prompts and phrasing here are our own, reviewed on 2026-10-09. Never paste confidential case details or client identities into a tool without checking your firm’s data policy, and treat every time and invoice figure the model returns as a draft that you verify against your billing system.