Can AI help us prepare a complaint to the Financial Ombudsman Service?

Yes, for the mechanical parts of the work. An AI assistant can read the client's file and the business's final response, build a chronology, set out the figures, and find published ombudsman decisions on similar facts. What it cannot do is decide what is fair and reasonable in your client's circumstances, which is the part the case turns on.

What this work actually consists of

A complaint to the Financial Ombudsman Service is usually less about law than about assembly and argument. The typical tasks are:

• Working out what the client was sold or told, and when.
• Checking the complaint has been through the business's own process and that the time limits in the regulator's rules have not run.
• Pulling the relevant documents out of a pack that often arrives as a single scanned bundle: application forms, suitability letters, illustrations, call notes, statements, the final response letter.
• Setting out what went wrong against the standards that applied at the time, including the regulator's rules and guidance, relevant codes and good industry practice.
• Quantifying the loss and saying how redress should be calculated.
• Anticipating what the business will say back.

Most of that is reading, locating and ordering. That is where an AI assistant earns its place.

Where AI genuinely helps

Reading the pack. Financial services files are long and repetitive. A tool that reads page by page and tells you which page a fact came from saves the slow work of hunting through a scan for the one paragraph in the suitability letter that matters. The useful question is not "summarise this" but "what does the file say about the client's attitude to risk, and on which pages".

Building the chronology. Dates drive these complaints: when the advice was given, when the client first knew something was wrong, when the complaint was made, when the final response was sent. A chronology drawn from the papers, with each entry pointing at the page it came from, is quick to check and quick to correct.

Finding comparable decisions. Published ombudsman decisions are not binding precedent, but they are the clearest available indication of how similar complaints have been approached. An AI assistant that reads those decisions in full, rather than matching keywords in a summary, can bring you three or four that are genuinely close on the facts. You still read them yourself before you cite them.

Working through the figures. Redress calculations involve several steps and each step rests on a rule or a published approach. A tool that shows the working, step by step, is far more useful than one that produces a number. You can check a working; you cannot check a number.

Drafting the complaint itself. Once the facts and the authorities are settled, a first draft of the submission is a reasonable thing to ask for. Treat it as a structured starting point, not a finished document.

Where it does not help, and where it can mislead

The ombudsman decides what is fair and reasonable in all the circumstances. That is a judgement, not a lookup. An AI tool has no view on whether this particular client, with this particular level of experience, should have understood this particular document. Asking it to predict the outcome is asking for a guess dressed as an answer.

Two further cautions.

First, standards change. What counted as good practice when the advice was given may not be what counts now. If a tool quotes guidance at you, check which version it is reading and whether that version applied at the relevant time.

Second, a confident tool with a thin source base is dangerous here. If it cannot find a decision on point, you want it to say so, not to invent one or stretch a distant one. Any tool that will not tell you when it has nothing is not safe for this work.

How to run it in practice

• Give the tool the whole pack, not a selection. If you pre-filter, you decide what matters before you have read it.
• Ask narrow questions with the answer tied to a page. "What did the final response say about the business's own file notes, and where" beats "is this a good complaint".
• Read every decision the tool brings you before relying on it. Open it at source and read it in full.
• Check the redress calculation line by line against the approach you would apply yourself.
• Keep your own diary of the deadlines. No tool should be the only record of when a complaint must be referred.
• Record on the file what the tool was used for and who checked it, in the same way you would record any other assistance.

The test for the fee earner is unchanged: you are putting your name to the submission, so you must be able to defend every date, every figure and every decision you have cited without reference to how the draft was produced.

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, and it reads official sources at the source, including Financial Ombudsman decisions in full, with every point naming its source and official sources opening in one click. 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. Where it cannot find support for a point, it says what is missing instead of guessing, and it prepares drafts for a qualified person to review and sign off.