London, 19 August 2026 – Financial advisers are open to agentic AI handling administrative work across the advice journey but want human oversight for decisions that affect client money, according to research from global wealth management technology provider GBST and consultancy the lang cat. The findings follow the Mills Review, published earlier this month, which recommended that the FCA monitor the shift towards systems making open-ended decisions without a person involved and adapt its regulatory framework accordingly.
The survey of 178 advisers found a strong appetite for the high-volume, rules-based tasks, with eight out of ten (80%) comfortable with AI collating data for annual reviews and suitability packs, 77% with onboarding and letters of authority, 76% with KYC and anti-money laundering checks and 75% with fees and charges reconciliation.
Comfort falls when AI moves closer to decisions that directly affect client money, though even here more advisers are comfortable or undecided than opposed. For pension transfers, 43% are comfortable, 29% neutral and 29% uncomfortable, while for CIP switching and rebalancing, 53% are comfortable, 22% neutral and 25% uncomfortable.
The research also found that familiarity with agentic AI is patchy. Nearly a third (31%) of advisers are unable to describe what it does, and a similar number (29%) cannot identify its core capability of breaking tasks into steps, planning the workflow and carrying it through automatically. However, the results suggest advisers may be more comfortable with the concept than the terminology. When shown a description of agentic AI, nearly two-thirds (62%) say they are comfortable with it being embedded in their platform, with a further 18% neutral.
The open responses also show clearly where advisers want the boundary to sit. Several said they wanted the ability to supervise and check outputs before anything is actioned, singling out tasks that directly affect a client as the ones they would not hand over without oversight. One adviser said: “Where there is a risk to client money, for now I think it is better for a human to handle it. A human will be more likely to spot if something is awry, whereas a computer will just push through regardless.” Others described their own move from caution to comfort. As one put it: “To begin with it feels risky, then you get used to it and feel comfortable, and you end up wondering how you would get on without it.”
Rob DeDominicis, CEO of GBST, said:
“Rather than resisting AI, advisers have drawn a sensible boundary around it. They are comfortable with agentic AI taking on the high-volume administrative tasks, like reconciliation and collation of data. This is necessary work, but it takes up time without adding visible value for clients. Where client money is directly at stake, they want human oversight, but that doesn’t mean doing everything manually. On more involved processes such as transfers, it’s about keeping people at the decision points while the system carries out the firm’s own procedure and records every step. The Mills Review makes the same distinction, recommending the FCA monitors this closely. Advisers have effectively drawn that line themselves, well ahead of any rule requiring them to.
“Advisers are most comfortable when AI is built into the platform they already use, rather than bolted on the side. That keeps it within the controls and security already in place. That’s how we’ve built Composer’s agentic AI and what we’re seeing from clients as these capabilities move from controlled release into live operational processes.”
