Canadian lawyers record time one of two ways: contemporaneously, with a timer running while the work happens, or by reconstructing the day afterwards from memory and files. AI does not invent a third method — it makes the first one survivable by proposing entries from work you have already done, which you then edit and approve.
Key takeaways
Contemporaneous docketing means the entry is created while the work is happening or immediately after. Reconstruction means sitting down at 6pm, or on Friday, and rebuilding the week from a calendar, a sent-items folder and memory. Almost every practice does some of both, and almost every practice knows which one produces the defensible record.
The problem with reconstruction is not honesty. It is that the small units disappear first — the six-minute call, the three-minute email that resolved a question, the twenty minutes spent reading a document before deciding it was not relevant. Those entries are the hardest to remember and the easiest to leave out.
The problem with contemporaneous docketing is that it interrupts the work. Starting a timer is a two-second task that a lawyer in the middle of a difficult paragraph will not do. That gap — between the discipline that produces a good record and the attention the work demands — is the actual problem AI is being pointed at.
Ontario licensees have to keep specified financial records. The Law Society of Ontario’s By-Law 9 lists them, and item 7 of that list is a fees book or a chronological file of copies of billings, showing all fees charged and other billings made to clients, the dates on which they were made, and which client was charged. Nothing in that requirement says how the underlying time was captured — only what the firm must be able to produce.
The fee that record supports has its own test. Rule 3.6-1 of the Rules of Professional Conduct says a lawyer must not charge or accept any amount for a fee or disbursement unless it is fair and reasonable and has been disclosed in a timely fashion, and the commentary lists the factors that make a fee fair — the first of which is the time and effort required and spent. Rule 3.6-3 adds that a statement of account must clearly and separately detail fees and disbursements.
British Columbia states the same principle in its own words: the Code of Professional Conduct for British Columbia rule 3.6-1 says a lawyer must not charge or accept a fee or disbursement, including interest, unless it is fair and reasonable and has been disclosed in a timely fashion. In Quebec, the Barreau du Québec notes that it publishes no directive telling lawyers which pricing model to use — the choice is the professional’s, subject to the Code of ethics of advocates.
Federally, the tax side reinforces it. Section 230 of the Income Tax Act requires records and books of account in a form that lets the tax payable be determined, and subsection 230(2.1) says that for a person carrying on business as a lawyer those records include all accounting records, including supporting vouchers and cheques.
Practice-management software has been narrowing the gap for years without any AI at all. Clio’s Canadian timekeeping product describes adding time entries directly from calendar events, tasks, communication logs, notes and documents, running a timer that can be paused, capturing time from Outlook or Gmail as messages are sent, and correcting an entry afterwards when a timer was left running or never started.
What generative AI adds is the narrative and the matching. An activity log knows that you spent 34 minutes in a document and sent two emails; it does not know that the work was "review of amended statement of defence and instructions to junior on reply". Drafting that sentence, in your firm’s house phrasing, and suggesting which matter it belongs to, is the part the model is good at. Clio describes its AI layer as converting time and expenses into finalised invoices ready for approval — note the last two words.
Treat every generated entry as a draft. The Office of the Privacy Commissioner points out in its principles for generative AI that the output of these systems is probabilistic and may vary even in response to the same prompt. A docket narrative is a statement to a client about what you did. It is not a place for a plausible sentence you did not read.
The figures below are an illustration rather than a measurement. Suppose a litigator has a Tuesday with four matters in it. Reconstructed on Friday, it comes out as 5.4 hours: a 2.0-hour block on the big file, 1.5 on a motion, an hour of "correspondence" and a 0.9 rounding of everything else. The three short calls and the two emails that actually resolved a scheduling dispute are gone, and so is the reading that preceded the motion.
Captured from activity and narrated the same evening, the same day comes out as twenty-one entries totalling 6.6 hours, of which the lawyer writes off 0.4 as duplicated effort and reallocates 0.3 to a matter that was billed on a fixed fee. The billable figure moved because the record improved, not because the day changed.
The second version is also the one that survives a challenge, because each entry names a specific act on a specific document at a specific time. That is what a client is entitled to see when they ask what the money bought.
First, where the data goes. Time capture reads calendars, email subject lines, document names and matter titles — which is client confidential information, and in most cases personal information as well. Rule 3.3-1 requires a lawyer to hold in strict confidence all information concerning the business and affairs of the client, and the fair information principles in Schedule 1 of PIPEDA apply to the personal information inside it. Get the data-handling terms in writing before the pilot, not after.
Second, whether anyone is being monitored. A tool that logs what staff do all day is an electronic monitoring tool whatever it is sold as, and in Ontario that carries a written-policy obligation for larger employers — covered in our article on capturing billable time without timesheets.
Third, what you are actually trying to fix. If the firm’s problem is that time is captured but never billed, better capture will not help; see AI for capturing billable hours instead.
Can AI write my docket narratives for me?
It can draft them from the underlying activity. You still read and correct each one, because the narrative is a representation to the client about work you performed.
Does automatic capture change what we can bill?
No. The fee still has to be fair and reasonable and disclosed in a timely fashion under rule 3.6-1. A more complete record can raise or lower the bill; it does not change the test.
Do we still need a fees book if the software has everything?
You need to be able to produce the records By-Law 9 lists. Software satisfying that is fine; software you cannot export from is not.
A 30-minute call is enough to tell you whether AI pays for itself here.
