Blog
Can AI help us prepare a letter of advice to a client?
Yes, for the groundwork: gathering the facts from the papers, finding the source material and structuring the letter. The advice itself, and the judgement behind it, stays with the fee earner.
Alesis · · 5 min read
Yes, for the parts of the job that are assembly rather than judgement. An AI assistant can pull the relevant facts out of the file with page references, find the legislation and official guidance you want to work from, and lay out a structure you can fill. What it cannot do is decide what the client should be told to do, and it cannot carry the professional responsibility for the letter that goes out under your name.
That distinction matters more in an advice letter than almost anywhere else. A letter of advice is the moment a client relies on you. It is also the document most often read back to you later, by the client, by a complaints handler, or by your insurer.
What the AI can usefully do
Most of the time spent on an advice letter is not spent advising. It is spent reconstructing what happened, checking what the documents actually say, and finding the source you half remember.
An AI assistant that works from the matter's own papers can help with:
- Setting out the facts as the file records them, with a page reference against each one, so you can see immediately where a fact came from and whether it is properly evidenced.
- Flagging where the papers are silent. If the file does not record when the client first knew of the problem, you need to know that before you write a paragraph that assumes it.
- Pulling the relevant provisions and official guidance at source, so you are reading the section rather than a summary of a summary.
- Producing a first structure: background, the question you were asked, what the documents show, the options, next steps, costs.
- Checking a later draft against the file, so that every factual assertion in the letter can be traced back to a document.
None of that is advice. All of it is work you would otherwise do by hand, usually under time pressure, usually at the end of the day.
What has to stay with the fee earner
The advice is yours. So is everything that shapes it.
You decide which facts matter and which are noise. You decide how much weight to give a document that is unsigned, undated or inconsistent with the client's account. You decide what risk appetite the client has and how to describe an uncertain outcome without either alarming them or giving false comfort. You decide what you are not advising on, and you say so in terms.
An AI assistant has no view on any of that. It has not met the client. It does not know what was said on the telephone last Tuesday unless someone wrote it down and put it on the file. It cannot tell you that the client's real concern is the business relationship rather than the money.
Treat AI output as material for the letter, never as the letter.
A workable order of work
A sequence that keeps the judgement in the right hands:
- Read the instructions yourself and decide what question you are answering. Write it down in one sentence.
- Ask the assistant to set out what the papers say on the facts that bear on that question, with page references.
- Read those pages. Not all of them, but the ones the answer turns on. If a page reference does not support what was said about it, that is the end of that point until you check further.
- Identify the source material you want, and read the provision or the guidance at source rather than relying on a paraphrase.
- Draft the analysis yourself, or dictate it, and use the assistant for structure, tightening and consistency checks.
- Check the finished letter against the file before it goes out.
The expensive mistake is reversing steps two and five: generating a polished draft first and then trying to verify it. Polished prose is persuasive. It is much harder to spot a missing fact in a fluent paragraph than in a bare list.
Checks before it goes out
Before an advice letter leaves the office, satisfy yourself on five things.
First, every fact stated is supported by something on the file, and you know where. Second, every source relied on has been read by a person, not summarised by a machine. Third, the scope is stated: what you have advised on and what you have not. Fourth, the assumptions are visible, particularly any fact the client has told you that you have not independently verified. Fifth, the letter answers the question the client actually asked.
Supervision applies in the ordinary way. If a junior prepared the letter with AI assistance, the reviewer is reviewing the letter, not the process, and should be reading the underlying documents on the points that matter.
Where Alesis fits
Alesis answers questions about a matter from that matter's own papers and names the page each answer came from; if the papers do not say, it says so. It reads legislation section by section and official guidance at source, and every point names its source, with official sources opening in one click. It prepares drafts for a qualified person to review and sign off, and it does not file, serve or email anything. It assists qualified professionals and does not replace them, and it does not provide legal advice.