AI may be transforming advice processes, but the quality of the output still depends on the quality of the information behind it. Steph Willcox, Head Actuary at Dynamic Planner, examines the role paraplanners play in maintaining confidence in the data that powers modern financial planning.
Everyone is racing to put AI into their day-to-day work. Claude is being rolled out into financial services roles in the US, and Dynamic Planner’s own research this year found AI adoption among UK advice firms jumped from 28% to 41% in a single year, with the 1% that had not yet introduced it expecting it to be useful in the near future.
While the rush to AI is on, none of it works without good data. A model can be excellent and still give you a bad answer if the numbers feeding it are wrong, incomplete, or simply not trusted by the person relying on them.
An AI tool summarising a client’s finances is only ever as good as the valuations, product data and assumptions sitting underneath it. Get those wrong, and a confident, well written summary is still a wrong answer.
While paraplanners have never been happier in the role, with 91% now saying they would recommend it as a career, the same research found paraplanners, of all the roles in a firm, are the least confident that they can get the data they need. The people who like their job most are also the ones closest to where the data breaks down.
Anyone who has pulled together a cash flow plan knows this problem in detail. A client’s pension valuation from one provider does not match the format of another, a protection policy needs manually checking against a schedule from three years ago, and a platform holding is there, but the figure attached to it does not quite tally with the statement in the file.
None of this shows up as an error but it causes an extra half hour, spent by someone quietly making the numbers usable before any planning conversation can happen.
Good integrations can fix the issue at source, to prevent all the re-work of the same information. When the mapping between a provider and a planning platform is solid and stable, a paraplanner does not need to re-earn trust in a figure every single time it appears.
It means the same thing this month as it did last month, in the same place, calculated the same way. That consistency is quiet and unglamorous, but it is what turns “the data is there” into “I believe this number” because access and trust are different things.
A system can technically hold the right numbers somewhere, and people will still redo the work by hand because they do not believe what it is telling them, or because getting to a trustworthy version takes real effort every time.
Ask most paraplanners whether the data is technically available and the answer is usually yes. Ask them whether they would hand it straight to a client without checking it first, and the answer changes.
A tool that drafts a suitability report, models a retirement income strategy, or flags an inconsistency in a client file is only trustworthy if the inputs feeding it are trustworthy first.
Firms are moving fast on the AI side of this equation but fewer are asking directly whether the data underneath has earned that same level of trust, or whether confidence has simply outpaced the work needed to deserve it.
So ask, do we trust the data we already have? If the answer is no, or even a hesitant maybe, that is the problem to fix before the clever model gets built on top of it.
The good news, highlighted in our data, is that the people best placed to answer that question honestly already know it and can be at the forefront of making it better.
Main image: boliviainteligente-DEci5GH0r0k-unsplash
