Can wealth managers use AI agents before staff are trained?

Yes, behind the scenes. A wealth manager can start agents preparing client review packs, reconciling holdings and drafting suitability reports while advisers and operations staff train. Recommendations, trades, transfers and acting on client instructions must wait until the people reviewing that work are trained, because the money involved is often a family's long-term security.

What makes wealth management different?

Wealth management combines regulated advice, large sums and long relationships. A suitability report that misstates a client's risk profile can lead to unsuitable advice. A transfer acted on from a spoofed email can empty an account. A client who is becoming frail may give instructions that do not reflect their real wishes.

Each of these failures passes through a person who could have caught it. An untrained reviewer approving an agent's draft is unlikely to.

What can agents do while staff train?

• Review packs from custodian and platform records: holdings, performance, fees, cash flows (depth 1).
• Portfolio drift reports against the agreed risk profile (depth 1 and 2).
• Suitability report drafts from the adviser's notes, for the adviser to edit and approve (depth 2).
• Fact-find updates drafted from a client's own submissions, marked as declared until confirmed (depth 2).
• Client correspondence drafts, such as meeting confirmations and valuation letters (depth 2).

What must wait?

• Recommendations of any kind.
• Placing trades or rebalancing portfolios (depth 4).
• Transfers and withdrawals, which should be held for a named person even after training.
• Acting on any instruction arriving by email or message, which is an external claim and cannot authorise a payment.
• Anything involving a client flagged as vulnerable.

What does a sensible start look like?

Take a wealth management firm with 20 advisers and an operations team of ten. It starts an agent drafting suitability reports from advisers' meeting notes. Advisers review each draft.

In the first month, the compliance officer samples approved reports and finds one that describes a client's attitude to risk as "adventurous". The adviser's notes said the client was adventurous in their twenties but now wants to protect capital. The agent had picked the wrong phrase; the adviser had skimmed the draft and signed it. The report is corrected before it reaches the client. The firm briefs advisers to read every draft against their own notes, and changes the brief so the agent quotes the client's stated attitude directly. All staff complete foundational training; advisers and senior operations staff work towards operator certification. Detail on planning is in the AI agent training plan template for wealth managers.

Which controls fit a wealth manager?

• The firm's own limits keep agents at depth 2 and hold every trade and transfer.
• Client instructions by email never act alone; they are verified through a known channel.
• Declared client information never treated as confirmed.
• Compliance samples approved suitability reports.
• Every verdict filed, so a complaint years later can be reconstructed.

Wealth firms often train advisers and forget operations, where transfer fraud is caught or missed, among the patterns in what wealth managers get wrong about AI agent training. The wider argument is in wealth management will not scale AI agents without trained reviewers. Banks face the same decision at greater scale, set out in whether banks can use agents before staff are trained, and so do utilities, in whether utilities can. A firm wanting an external standard for its reviewers can use Pryme Intelligence Academy's Operator certification. The topic guide on training teams for AI agents sets out the order.