Blog
How should a team of fee earners use AI on the same matter?
Keep the AI work for a matter in one shared place rather than scattered across individual accounts, agree who may ask and who must check, and put the conclusions on the file rather than leaving them in a chat history.
Alesis · · 5 min read
If more than one person is working on a matter, the AI work should sit with the matter and not with the individual. That means one shared thread or conversation per matter, access limited to the people actually working on it, and the conclusions written up on the file in the normal way. The chat is a working tool; the file remains the record.
Keep one thread per matter, not one per person
The common failure mode in a team is duplication. The partner asks the tool to summarise the expert's report on Monday. The associate asks the same question on Wednesday, phrased differently, and gets an answer with a different emphasis. The paralegal, who has seen neither, asks a third time on Friday. Nobody knows which version informed the letter that went out.
A single conversation per matter avoids most of this. Everyone can see what has already been asked, what came back, and what was rejected. Questions build on each other rather than starting from nothing. If someone has already established that the lease is silent on a point, the next person does not spend twenty minutes rediscovering it.
It also makes supervision possible. A supervisor who can read the thread can see how the matter has been interrogated, not just what the final answer was.
Decide who has access, and on what basis
Access should follow the matter, not the org chart. The people working on the file can see the file; everyone else cannot, regardless of seniority. That is the same principle most firms already apply to their document management system, and it should apply to any AI tool that holds matter papers.
Things to settle before the team grows:
- Who adds and removes people when the team changes, and how quickly it happens when someone moves department or leaves.
- Whether support staff who scan and index documents need access to the conversation, or only to the documents.
- How a conflict or an information barrier is handled. If a barrier is in place on the matter, the AI tool has to respect it like any other system.
- Whether anyone outside the team, a supervisor or a COLP carrying out a file review, needs read access and on what footing.
Write the answers down once. They will be the same on every matter.
Agree who asks and who checks
The most useful rule a team can adopt is that the person who asks the question is not automatically the person who decides the answer is right. On a matter with a trainee, an associate and a partner, the trainee may do most of the asking. The checking still belongs to whoever is accountable for the work product.
In practice that means three habits:
- The person asking states what they were trying to establish, not just what they found. "I was checking whether the notice period runs from receipt or from posting" tells a supervisor far more than a pasted answer.
- Every point that will go into advice, a pleading or a letter is checked against the underlying source. If a tool names a page or a section, open it. If it does not name a source, treat the point as unverified.
- Nothing goes out on the strength of the AI answer alone. A qualified person signs off, as they would with any junior's draft.
Set out in the team's opening email who holds the pen on each of these. On a small matter that may be two people. On a larger one it is worth saying it explicitly.
Make the file, not the chat, the record
A conversation thread is a working space. It contains false starts, rephrased questions and answers that were later abandoned. It is not a substitute for an attendance note or a research note, and it should not be left to do that job.
When a question is resolved, write the conclusion on the file in the usual place, with the source relied on and the date. Then anyone picking up the matter in six months, including a costs assessor, a new fee earner or an insurer, can see what was decided and why without reading a transcript.
The same applies to key dates and calculations. If a tool has worked out a deadline, the date goes into the diary system and the working goes on the file. The diary governs. No team should be relying on a chat thread to remind it of a limitation date.
When to start again
There are a few points at which a fresh start is better than carrying on:
- A large new tranche of documents arrives and changes the picture materially.
- The matter splits, for example into separate proceedings against different parties.
- The instructions change so fundamentally that earlier questions are now about a different case.
In each case, say in the file note why the thread was closed and where the next one begins. Continuity matters more than tidiness, but a thread that has drifted across three different disputes helps nobody.
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. Inside a firm, people see only the matters they are on, and seniority alone grants no view. It answers questions about a matter from the matter's own papers and names the page each answer came from, so a colleague picking the thread up can check the source rather than take the answer on trust. It prepares drafts for a qualified person to review and sign off, and it says that the firm's diary system still governs.
