Can AI help us deal with the day's incoming post on a busy caseload?

Yes, within limits. An AI assistant can read the day's letters, emails and enclosures on a matter and tell you what each item says, what it asks for, what dates it raises and whether it is consistent with what is already on the file. What it cannot do is decide what any of it means for the case, reply to it, diarise it or close it off. That part is fee earner work, and it stays that way.

What post triage is actually asking

When a fee earner opens the day's post on twenty or thirty live files, each item raises the same three questions:

1. What is this, and does it need a response?
2. Does it change anything: a date, a figure, an assumption, an instruction?
3. Who needs to know, and by when?

The difficulty is rarely that the questions are hard. It is that there are a lot of them, they arrive in no useful order, and some of the answers depend on remembering what a letter from four months ago said. That is the part where a tool that has read the file is genuinely useful.

What AI handles well on incoming correspondence

• Compressing a long letter into its asks. A six page letter from the other side often contains three actual requests buried in commentary. Asking for the requests, with the page each one came from, takes a minute and gives you something you can act on.
• Pulling out dates and figures. A letter that says a response is required within a given period, or that attaches a revised schedule, can be reduced to a short list of dates and numbers with the working shown.
• Checking consistency with the file. "Does this letter match what they told us in the earlier correspondence about the inspection?" is a question that is quick for a tool that has the papers and slow for a human who has not read them this week.
• Spotting missing enclosures. If a letter refers to an appendix that is not there, that is worth catching on the day rather than a fortnight later.
• Flagging what it could not read. A poor scan or a handwritten note should be flagged as unread, not quietly skipped. If a tool does not tell you what it could not read, you cannot trust the summary.

None of this is glamorous. It is the reading, and the reading is most of the time cost.

Where judgement stays with the fee earner

Triage is not only comprehension. It is also tactics, and a tool working from the papers alone has none of the context that makes a decision sensible.

It does not know that the client has already said they want this settled by the end of the quarter, unless that is on the file. It does not know that the other side's solicitor is reliable about extensions, or that counsel is unavailable in the relevant week. It cannot weigh whether a request is reasonable, whether responding engages a costs risk, or whether an apparently routine letter is the first sign of a problem.

So the division of labour is simple. The tool answers what the papers say. The fee earner decides what to do. Where a summary looks like it is drifting from description into advice, treat that as a prompt to go and read the letter yourself.

A routine that works in a real morning

The firms that get value from this tend to standardise it rather than improvise:

1. Work matter by matter, not inbox by inbox. One conversation per file keeps the context where it belongs.
2. Add the day's items to that matter, including enclosures.
3. Ask the same three standing questions each time: what does this ask for, what dates or figures does it raise, and does it contradict anything already on the file.
4. Check each answer against the page it came from before you act on it. A citation that points at a page is only useful if someone looks.
5. Put every date into the firm's diary or case management system yourself. A calculation from a tool is a prompt to diarise, never the diary entry.
6. Record the outcome on the file in the usual way, including anything you decided not to do and why.

Risks worth naming before you start

The main one is assuming the tool has seen more than it has. If half the correspondence arrived by phone or came in on a colleague's email, a summary will be confident and incomplete. Second, watch for items that contain a third party's personal data, medical records or material belonging to another client; the usual confidentiality and data protection thinking applies, and your supplier's terms matter. Third, a summary is not a substitute for reading anything that is contentious, unusual or likely to be read back to you later.

Used carefully, this is one of the lower risk uses of AI in a firm, because every output can be checked against a document you already hold.

Where Alesis fits

Alesis is an AI assistant for UK law firms, used through the web browser, with one conversation for a matter. It answers questions about that matter from the matter's own papers and names the page each answer came from; where the papers do not say, it says so, and any page it could not read is flagged rather than skipped. It counts key dates and figures with the working shown, each step carrying the rule that allows it, and says the firm's diary system still governs. It prepares drafts for a qualified person to review and sign off; it does not file, serve or email anything.