AI and deadline calculation: why you still need a rules-based diary
By the ChamberOne team 5 minute read
Of all the mistakes a lawyer can make, a missed deadline is among the least forgivable. It is also one of the easiest to make with an AI tool, because the question sounds simple: what is 28 days from this date? A language model will answer immediately and with confidence. It may even be right. But it does not count the way a calendar does, and the cost of a wrong answer is out of all proportion to the effort of checking.
Key points
- Language models predict plausible answers. They do not run a calendar, so they can miscount days and holidays.
- Use rules-based software or a documented method to calculate dates, and keep AI for reading and flagging.
- Have a second person verify limitation dates and any date that would be serious to miss.
- Record the trigger, the rule and the source of each date on the matter.
Why language models and dates do not mix well
A language model produces the words that are likely to follow your question. It does not run a calendar. When it adds days, it is predicting a plausible answer, not carrying out a calculation. For small, common sums it is usually right. For longer spans, leap years, month ends and awkward cases it can slip, and it will not tell you it has.
Dates also depend on facts and rules that the model may not have: which day is the starting point, whether it is included, whether weekends and holidays count, what happens if the last day falls on a closed day and which holidays apply in a given court or place. Each of those varies, and the model may blend the practices of different places.
Where the real risk lies
Arithmetic errors are one risk. The bigger one is the rule behind the arithmetic. Limitation periods start on different events in different cases. Some can be extended or suspended. Procedural time limits may be set by statute, by court rules, by a practice direction or by an order made in the case. Orders are often specific and unusual, and a tool will not know them unless you tell it.
A summary tool may read a court order and give the wrong date, or fail to notice that a deadline has been varied by a later order. It may confuse the date of an order with the date of service. Those errors come from interpretation, not counting, and rules-based software does not fix them either.
What a rules-based diary does
A deterministic tool applies explicit rules the same way every time. A diary system with court rules built in will take a trigger date and a type of step, apply the counting rules and holiday calendar for that court and produce a date. If you run it twice, you get the same answer. If a rule is wrong, it is wrong in a way someone can find and correct.
That predictability is the point. It means the answer can be reviewed, tested and trusted in a way that a probabilistic answer cannot.
- Predicts a plausible answer rather than running a calendar
- Can slip on long spans, leap years and month ends
- May blend the practices of different places
- Will not tell you when it has slipped
- Applies explicit rules the same way every time
- Uses the counting rules and holiday calendar for that court
- Gives the same answer if run twice
- A wrong rule can be found and corrected
A sensible division of labour
- 1
Use AI to read a document and propose the trigger event and the type of deadline, flagged for a person to confirm.
- 2
Use a rules-based tool or a documented manual method to calculate the date.
- 3
Have a second person verify any limitation date and any date that, if missed, would be serious.
- 4
Record the source of the date and the rule used, on the matter.
- 5
Diary reminders well before the deadline, not only on the day.
Verification habits
Whatever the source of a date, treat it as unverified until it has been checked. A few habits catch most errors.
- Count independently, using a calendar, for any date that matters.
- Check the rule that sets the period against the primary source, not a summary.
- Confirm the starting event and its date from the document itself.
- Check for later orders or amendments that vary the period.
- Allow a margin of safety: aim to act well before the last possible day.
Limitation is its own discipline
Limitation periods deserve special treatment. They often turn on when a cause of action arose, when damage occurred or when something was discovered, and on the claimant's age or capacity. They may be subject to exceptions. Take any AI summary of limitation as a prompt for research, not an answer. Have a senior lawyer sign off the limitation date on every contentious matter and record it prominently on the file. The best practice is to treat limitation as a separate check, with its own diary entry and an owner.
Use AI to build the checklist, not to own the date
Where AI does help is in the surrounding work. It can extract all the dates mentioned in an order and list them for review. It can compare two orders and point out that the later one changes a deadline. It can draft the checklist of steps a procedure involves. Each of those outputs is a suggestion, and the lawyer or the diary tool still owns the actual date.
Suppose a litigation team uploads a new directions order. The tool lists five dates. A paralegal enters each into the diary system, which calculates reminders. A solicitor checks the five against the order itself, initialling each. The tool saved time on reading. The diary system and the solicitor provided the certainty.
Spreadsheets and chatbots are not diaries
Many firms rely on a spreadsheet of deadlines kept by one person. That is a rules-free system with a single point of failure, and swapping the spreadsheet for a chatbot does not improve it. Whatever you use, the system should show the rule that produced each date, who entered it, who checked it and when it was last changed. If you cannot see those things, you cannot audit the diary, and you will find out how good it is only on the day something goes wrong.
- Use AI to extract dates from an order and compare orders
- Let the diary tool or lawyer own the actual date
- Show the rule, who entered it and who checked it
- Leave the actual date to a chatbot
- Rely on one person's spreadsheet of deadlines
When you must trust a date
Sometimes a deadline is too close to allow elaborate verification. In that case, choose the earlier of two possible dates, act on it and clarify afterwards. A filing that is early costs nothing. A filing that is late may cost the claim. Keep a record of how the date was derived, so that if anything goes wrong, the file shows a reasonable process.
This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.
Read next
Onboarding a new client matter: an AI-assisted step-by-step checklist
Walk through every step of opening a new matter — from conflict check to task assignment — showing where AI saves time and where only a lawyer can decide.
6 minute readUsing AI matter briefs to pick up an unfamiliar file quickly
How a generated brief speeds up handovers and cover, and how to check it against the file before acting.
5 minute readAI task suggestions and matter phases: standardising how work starts
How matter-type templates and suggested tasks help new files start consistently, and why each suggestion needs review.
5 minute read