What changes
Most advice firms know their fee income to the pound and their adviser time only roughly. Taking stock fixes that. You list everything the firm does for an ongoing client, from the first meeting to the annual review, and note who does each task and how long it takes. Then you ask one question of every task: could an AI tool do most of this today, with a person checking the result?
Next, score the firm against the six principles of the New Model. Give each a mark from 0 to 3, where 0 means nothing is in place and 3 means it is part of how the firm works every week. The scores show where to start. A firm that scores well on relationships but has no AI policy starts with Move 2. A firm with capable staff and a fee model built on assets under management may find that Move 5 needs the longest run-up, so it plans for it early.
Why it matters now
AI tools can already draft a suitability report, summarise a meeting and pull together a review pack in minutes. If capability keeps improving at anything like the recent pace while staying cheap to use, more of the work behind an adviser fee will take a fraction of the time it does today. The value of advice sits increasingly in judgement, the relationship and accountability. You can only price around those once you know how much of your current fee pays for work AI can now carry.
There is a regulatory reason too. Under the FCA Consumer Duty, firms must be able to show that the price a client pays is reasonable for the benefit they receive. The FCA has also asked firms to check that clients paying for ongoing advice actually receive the ongoing service. A clear map of what each client gets, and what it costs you to provide, is the evidence both questions need.
Checklist
- List every recurring task for a typical ongoing client: review preparation, the meeting itself, file notes, suitability letters, rebalancing, valuations, chasing providers and answering client emails.
- Time each task for two weeks. Ask advisers, paraplanners and administrators to log real minutes against each task rather than giving estimates at the end.
- Mark each task as one AI can do most of now, one AI can help with, or one that needs a person. Test at least five of them on a current AI tool using anonymised client data.
- Cost the service by client band using your true cost per hour, and set that figure beside the fee each band pays today.
- Score the firm 0 to 3 on each of the six principles as a team, then agree the one move you will start this quarter and who owns it.
Illustrative example
A three-adviser firm has 300 ongoing clients. Its two-week time log shows each annual review takes about 6 hours in total: 1.5 hours in the meeting and 4.5 hours of preparation, file notes, letters and admin. Across the client bank, that’s 1,800 hours a year.
The team tests review preparation and file notes on an AI tool, with a paraplanner checking every output. They estimate the 4.5 hours of surrounding work could fall to about 2 hours. That would free 2.5 hours per client, or 750 hours a year, which is roughly half of one adviser’s working year. The firm scores itself 2 on relationships, 1 on AI capability and 0 on its own digital offering, so it starts with Move 2 and plans Move 3 alongside it.
Common mistakes
- Estimating time instead of logging it. Guesses made after the event tend to understate admin, so the costing that follows comes out wrong and the fee work in Move 5 starts on a weak base.
- Scoring the firm generously. A score of 3 means it happens every week without anyone pushing. If it depends on one enthusiast, it is a 1 or a 2.
- Testing AI on made-up cases. Use real, anonymised files so you see how the tools cope with the untidy detail of actual clients and their paperwork.