How can AI help a fee earner get on top of an inherited matter?

Taking over a matter from a departing colleague, another firm or a long-dormant file usually means reading a lot of paper before you can do anything useful. AI can shorten the mapping stage: it can pull out a chronology, list the parties and the documents, surface the dates and figures, and tell you which questions the papers do not answer. It cannot tell you what the matter is worth, what the client actually wants, or which of the previous fee earner's decisions were sound.

What the first pass should produce

The purpose of the first pass is not to understand the matter. It is to know the shape of it well enough to decide where to spend your reading time.

A useful first pass gives you:

1. A chronology of events with a page reference against each entry, so you can go straight to the source.
2. A list of the people and entities named, and how they are described.
3. Every date that looks like a deadline, a limitation point, a contractual notice period or a court direction.
4. Every figure that looks like a claim value, a payment, an interest rate or a costs estimate.
5. A list of the documents themselves: what is on the file, in what form, and what is referred to but not present.

That last point is the one firms undervalue. The gaps in an inherited file are usually more urgent than the contents. A letter that refers to an enclosure you cannot find, a witness statement in draft with no signed version, an order referred to in correspondence but not on the file: these are the things that cost time later.

A sensible order of work

Start with the deadlines. Before you read anything for meaning, find out whether anything is due. Ask for every date in the papers and what it appears to relate to, then check each one yourself against the primary document and against the firm's diary system. If a limitation point is anywhere near, that is a partner's problem within the hour, not a matter for your reading list.

Then build the chronology. Ask for it in date order with page references, and read it as a table of contents rather than as a narrative. Where two documents conflict on a date, you want to know that they conflict, not to be given a tidy version that hides it.

Then read properly. Pick the ten or fifteen documents that the chronology suggests actually drive the matter and read them yourself, end to end. This is not optional and it is not a formality. The judgement you are being paid for comes from the primary documents, and a summary of a contract is not a contract.

Finally, write your own note. A short handover note in your own words, listing what you know, what you have assumed and what you need to ask the client, is the artefact that makes the file yours. Produce it after you have read, not before.

How to ask questions of a file you do not know

Broad questions on an unfamiliar matter produce broad answers, and broad answers on a file you cannot yet sanity-check are the most dangerous kind.

Ask narrow, checkable questions:

• "What does the file say about when the defendant was first notified?"
• "Which documents mention the retention of title clause?"
• "Is there a signed version of the settlement agreement on the file?"
• "What is the last correspondence with the client, and what did we tell them?"

Each of these has an answer you can verify by opening one page. That is the point. On a file you know well you can spot a wrong answer from experience; on an inherited file you cannot, so every answer needs to point somewhere you can look.

Be equally interested in "the papers do not say". An honest absence is a piece of information. A confident answer assembled from general knowledge about how matters of that type usually go is worse than useless, because it will read as though it came from the file.

Where the risk sits on an inherited file

Three risks are specific to takeover work.

The first is inherited error. If the previous fee earner made a mistake, summarising their file faithfully reproduces the mistake. AI reads what is there; it does not know what should have been there. Someone senior needs to ask whether the steps taken so far were the right ones.

The second is missing context. Telephone attendance notes, informal agreements and things the client said in a meeting may never have reached the file. Assume there is context you do not have and ask the client early.

The third is scanned material. Older files often contain poor scans, faxes and handwriting. A tool that quietly skips a page it could not read leaves a hole you will not notice. You want to be told which pages were unreadable so you can deal with them by hand.

None of this changes your professional obligations. The work is yours once you take it, whatever state it arrived in, and the client is entitled to the same standard of care and the same clear communication as if you had opened the file yourself.

Where Alesis fits

Alesis is an AI assistant for UK law firms, made by L25 Limited, used through the web browser, with one conversation for a matter. It answers questions about a matter from the matter's own papers and names the page each answer came from; if the papers do not say, it says so. Documents are read page by page, so citations point at pages, and any page it could not read is flagged rather than skipped. It counts key dates and figures with the working shown, and says the firm's diary system still governs.