Judge AI suitability report software by whether the client finishes the report knowing why

Anthony Marris argues that AI suitability report software should be judged by the client's experience of the finished report, because Consumer Duty asks firms to show a report is clear enough for the client to act on. He sets out how AI drafting can repeat the patterns that lose clients, and the plain checks a firm can run on a real file before buying: clarity of the recommendation, a readable disadvantages section, sensible length, figures that match the firm's records, and handling of vulnerable clients.

A 30-page report that nobody reads is less useful than a 10-page report the client actually engages with. That line sits in our guide to writing a suitability report, and it's the most useful thing I'd carry into any demo of AI suitability report software.

Most of the buying conversation about these tools is about the firm. How long a draft takes. How much paraplanner time it saves. Whether compliance will be comfortable. All fair questions. But a suitability report has two readers, and the guide is clear about the order: the client comes first, and the compliance reviewer reading it after the fact is a secondary audience.

So I'd judge an AI drafting tool by a simpler outcome. Does the client finish the report knowing what you've recommended, why it suits them, and what they're giving up?

The duty already sets the bar

Consumer Duty put this in writing. The guide points to the consumer understanding outcome as the one that bites hardest on drafting: the firm has to evidence that the report communicates clearly enough for the client to act on it. A report can be technically complete and still fall short.

That's a high bar for a human writer, never mind a model. And AI cuts both ways here. A model that's good at prose can produce a report that reads beautifully and says very little about this client in particular. It can also turn a short, clear recommendation into twelve pages of tidy filler, because length costs a machine nothing.

Cheap for the firm. Expensive for the client.

What the client feels when a draft misses

Think about what a client actually does with the report. They read the first page properly. They skim the middle. They come back to it months later, when a market wobble makes them nervous or when a partner asks why the pensions moved.

That's when a weak report shows itself. The guide lists the reasons reports come back from compliance, and read from the client's side it's a list of the places a client gets lost. Objectives written as "plan for retirement" instead of "retire at 62 with £42,000 a year in today's money". Costs given as a percentage with no cash figure. A disadvantages section that reads like product literature, when it should be telling this person what they're accepting. Reasoning that explains why the firm's portfolio suits a segment, with nothing on why it suits them.

Each of those is a pattern an AI tool can repeat at speed if it's built around templates instead of the client's own record. The client won't spot the template. They'll just feel, vaguely, that the document isn't really about them, and they'll ring the adviser with the question the report should have answered.

What I'd check before buying

None of this needs a technical test. It needs a real case and a fresh pair of eyes.

Take a recent file, run it through the tool, and give the draft to someone in the firm who has never seen the fact-find. Ask them to read the recommendation section and tell you, in one sentence, what's being recommended. The guide says a client should know that from the first two sentences of the section. If your reader can't, the client won't either.

Then try the disadvantages section. The guide's test is whether a non-technical reader could finish it and accurately summarise what they're accepting. Ten minutes. It tells you more than a feature list.

Look at length next. Does the tool let you keep the report short, pointing back to the fact-find instead of restating it? The guide suggests anchoring the reading age at 16. Ask whether the tool can check that, or at least doesn't fight you when you cut a page.

Check that the numbers survive the plain English. When a draft says a 25% fall still leaves retirement income above the client's floor, every figure in that sentence should be the one in your records, and the all-in cost over the horizon should be there in pounds. Friendly wording that drifts from the facts is worse than dry wording that's right.

And ask about vulnerable clients. The guide treats vulnerability indicators in the fact-find with no adjusted communication as a rejection reason in its own right. How does the tool change its drafting when the record flags vulnerability, and does it hand those cases back to a person?

Where the saved time should go

Wealth Analytica builds AI drafting into its suitability work, so I'm not neutral on whether firms should use these tools. I think they should. But the time a tool saves is only worth having if some of it goes back into the report: the extra pass that cuts a page, the objective in the client's own words, the cash figure next to the percentage.

Clients rarely tell you a report was too long. They just stop reading, and the adviser finds out at the next review, when the client asks something the report already covered.

If you handed your last suitability report to the client's partner, could they tell you why the advice suits the client?

Every Wealth Analytica article is fact-checked against primary sources where applicable. Read our editorial policy for our sourcing and review standards.