Wealth management will not scale AI agents without trained reviewers

Wealth firms want agents to let each adviser serve more clients. That only happens if the paraplanners, client service staff and advisers who review the agent's work are trained to judge it. Suitability drafts, rebalancing proposals and withdrawals all stop at a person, so untrained reviewers turn the agent into one more thing to check.

What does an agent actually do in a wealth firm?

Most of the value sits in preparation and routine servicing. An agent assembles a meeting pack from the client file and portfolio data at depth 1. It drafts a suitability report or an annual review letter at depth 2, which a paraplanner checks and an adviser signs. It proposes rebalancing trades to bring a portfolio back within its model tolerance at depth 3, executing those that stay inside the firm's limits and holding the rest. A client withdrawal or a transfer to another provider moves money and is hard to undo, so it sits at depth 4 and waits for a named person.

Advice itself stays with the adviser. The agent can lay out the client's stated objectives, the risk profile on file and the products that fit, but the recommendation and its accountability belong to a qualified person.

Why do reviewers set the ceiling on scale?

A wealth firm scales by adding clients per adviser without lowering the quality of advice. Agents help only if the review of their output is faster than doing the work, and still careful.

That balance breaks in two familiar ways. An adviser with a full diary signs a suitability report the agent drafted because it reads well, without checking that the client's capacity for loss was updated after a recent change in circumstances. Or a cautious paraplanner rewrites every draft from scratch, so the agent saves nothing. The first creates conduct risk. The second creates cost. Neither scales.

The pattern is not unique to wealth. Firms that look after other people's assets meet it too, as property management will not scale AI agents without trained reviewers shows for landlords' money and tenants' deposits.

Which decisions need the most skill?

Three, each specific to wealth work.

Client instructions from outside channels. A request to withdraw funds that arrives by email is external: it cannot authorise a payment on its own, however familiar the sender looks. Reviewers must know to confirm through a verified channel before releasing anything, and that the agent's held verdict is there to make them do so.

Suitability drafts. The reviewer must check the facts the draft relies on and their provenance. A risk score from the firm's own profiling system is sealed. Income or objectives the client stated in a meeting are declared. The draft is only as sound as the weakest fact it leans on.

Held rebalancing trades. When a trade is held because it breaches a concentration limit or a client restriction, such as an ethical exclusion, the reviewer decides whether to release, amend or refuse it, and writes the reason down.

How do reviewers affect the agent's depth?

Depth is earned, not configured. A firm may want its rebalancing agent at depth 3 across every model portfolio. Attest decides by replaying the agent against real cases from the firm's own book and scoring each answer against a floor, and those cases carry the outcomes reviewers chose.

If reviewers approved trades that breached a client's restrictions, or refused sound ones for comfort, the evidence is muddied and the agent stays at drafting. Certificates lapse after 90 days, and a run that clears no depth drops the agent to depth 1 the same day, so the quality of review is tested again at every renewal, not just at launch.

In the UK, the FCA's Consumer Duty expects firms to deliver good outcomes for retail clients, and the Senior Managers and Certification Regime makes named individuals accountable. In the US, the SEC and FINRA supervise advice and brokerage. None of these lets a firm say the agent made the decision. This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.

How do you measure whether reviewers are ready?

Use a proficiency scale with plain levels per skill: aware, supervised, independent, able to teach. A paraplanner might be independent on suitability drafts and only supervised on held withdrawals. Grant approval rights to match, so nobody releases money at depth 4 until they are rated independent on that decision.

Then assess on real cases, not quizzes alone. Take a run of the agent's actual drafts and held actions, have each reviewer decide them, and mark the decisions for correctness and for any missed red line. A firm can adopt the practice of tying approval rights to a current certificate, depth by depth.

Can a firm start before its people are trained?

At depth 1 and 2, with care: meeting packs and draft letters that a person must send are low risk and make good training material. The fuller answer is in whether wealth managers can use AI agents before staff are trained. Property firms face a similar question with tenant money, set out in whether property management firms can use AI agents before staff are trained.

What a firm should not do is let an agent execute rebalancing or prepare withdrawals for release before the people deciding its held actions have been assessed. The sequence that avoids this is in the guide to training teams for AI agents.

Where the Academy fits

Pryme Intelligence Academy trains advisers and directors through the Foundations track, about four hours, and certifies paraplanners, client service staff and team leads through the Operator track. Operator certification needs a written exam and two weeks of real review decisions with no missed red line, signed off by a Pryme assessor. Certificates are issued through VerifyDoc.ai and checkable at source.