Lawyers lose real money to time they worked and never recorded. AI closes the gap two ways: passive capture, where software watches what you were doing and drafts the entries, and entry writing, where AI turns “call w/ client re: settlement” into a narrative a client will pay without arguing.
The leakage almost never happens at the desk. It happens on Thursday afternoon when you sit down to reconstruct Monday.
Why reconstructed time is always short time
Time recorded days later is time under-recorded. That is the whole problem in one sentence.
Think about what a real day looks like. Nine minutes reading a discovery response. Six minutes on a call you did not schedule. Four minutes drafting a two-line email that required you to reread a contract clause. Twelve minutes on a docket question from an associate. None of those get written down when they happen, and none of them survive a memory sweep at the end of the week. You remember the deposition and the brief. You do not remember the fourteen small things, and the fourteen small things were two hours.
I am not going to quote you a statistic here, because the widely circulated numbers on this come from vendor surveys and I do not trust them. You do not need a study anyway. Run the experiment on yourself for two weeks and you will have a number that is actually about your practice.
Passive capture tools
Passive capture means the software runs in the background, logs what applications and documents and calls you touched, and hands you a draft timesheet to approve. You edit instead of remember.
| Tool | How it captures | Price range | Best for |
|---|---|---|---|
| Rize | Watches active app and document activity, auto-categorizes into projects | About $10 to $17/month | Solos who want the most complete raw record |
| Timely | Logs activity, then AI drafts the timesheet entries for you | About $11 to $25/user/month | Small firms wanting drafted entries, not just logs |
| Toggl Track | Manual and timer-first, with strong reporting and utilization views | Free to about $9/user/month | Firms whose problem is reporting, not capture |
| Clio / MyCase built-in timers | Timers plus AI assist inside the billing system itself | Included in the practice management plan | Firms that want entries to land where they bill from |
The tradeoff is straightforward. Standalone tools like Rize and Timely capture more, but you move the entries into your billing system. Built-in timers capture less, but nothing gets lost in transit. If your firm already lives inside a practice management platform, start with what is in the box and only leave if it fails the two-week test below.
One warning about passive capture: it sees everything, including matters and documents that are privileged or personal. Read the vendor’s data policy before you install, and keep the local-processing option on if the tool offers one.
The other half: descriptions that survive review
Captured time still gets written off if the description is bad. “Attention to file, 1.4” is an invitation for a client to cut it. Clients pay entries that tell them what happened and why it mattered to them.
This is what AI is genuinely good at. Paste in a day of shorthand, get back clean narratives, read them, send them.
You are a billing assistant for a law firm. Convert my shorthand time
entries into client-facing billing narratives.
Rules:
- Active voice, past tense. Start with the verb.
- State the specific work done and the benefit to the client.
- No block billing. One task per entry. Split any entry containing
multiple distinct tasks and split the time proportionally.
- 12 to 25 words per entry. No filler like "attention to" or
"various matters."
- Use the exact time I gave you. Never round up, never inflate,
never add time.
- Do not name AI tools in the narrative.
- If an entry is too vague to describe honestly, return it flagged
as NEEDS INPUT with a question instead of guessing.
Format: date | time | narrative
Here are my entries:
[paste]The last rule is the one people skip. If you let AI fill in what it does not know, it will write a plausible narrative for work you may not have done, and you will sign your name to it.
The same principle applies to the raw material. If your notes come out of an AI notetaker sitting on your calls, or out of dictation between meetings, the shorthand is already captured and the narrative step becomes almost free.
The ethics line, and it is a bright one
Model Rule 1.5 requires a reasonable fee. ABA Formal Opinion 512 addressed generative AI and fees directly.
Here is the rule in practice. If a task used to take you an hour and AI now does it in three minutes, on an hourly matter you bill three minutes. Not the hour. Not some blended “value of the work” number. The efficiency belongs to the client, because the client is paying for your time and your time is what got shorter.
You also do not bill machine time as attorney time. If a tool ran for 40 minutes summarizing a deposition while you did something else, that is not 40 minutes of your work. Bill the time you spent directing and reviewing it. And if you are passing through a subscription cost, disclose it and get agreement in the engagement letter first.
None of this is a reason to slow down your adoption. It is a reason to change your pricing model.
The flat fee argument
Flat fees get better the more AI you use, and hourly fees get worse.
Under hourly, every hour AI saves you is revenue you delete. Under a flat fee, you priced the outcome, the client agreed to it, and compressing the labor is your gain and the client’s certainty. That is legitimate value pricing, and it is the only clean way to capture the upside of moving faster.
The firms I watch struggling with this are the ones trying to keep hourly billing while quietly benefiting from AI speed. That does not hold. Pick one.
How to implement this in two weeks
- Pick one capture method. Built-in timer if you live in Clio or MyCase, Rize or Timely if you want maximum capture.
- Run it in parallel with your current process for two full weeks. Do not change how you normally record time. You need the comparison.
- Compare the totals. Recorded hours under your old process against what the tool captured, matter by matter.
- Write your narratives with the prompt above for one billing cycle, and read every line before it goes out.
- Take the three matter types where the gap was largest and quote them flat next quarter.
Start step one today: turn on a capture tool before your next call, and let it run while you keep timing the way you always have. In two weeks you will know your own number instead of somebody’s survey.