Ten Minutes on Decision AI, an Hour on Trust

The first wave of AI in wealth management has been about language - drafts, summaries, copilots, meeting notes. The next wave will be about decisions. Risk. Suitability. Portfolios. The advice itself.

We thought we'd spend two hours arguing for it, but many wealth firms have already done the language-AI work. Their boards are asking them what comes next, and they know full well it isn't another copilot. We’ve moved beyond the hype phase of AI, and we’re fully in the heavy work of how to make it valuable every day in business environments.  The argument we'd anticipated was an argument they had already been making to themselves.

So the discussion advanced quickly, and what came next wasn't about the technology. It was a conversation about trust.

Dr Nadine Kroher of Passion Labs, who co-hosted the evening with us, sparked an interesting inflection point by making the technical case that AI can reduce bias in financial advice.

Reinforcement-learning models can simulate hundreds of thousands of scenarios, find the strategies most likely to meet a client's goals, and show their working. "I want to see numbers," she said. Her case was the one a scientist is most likely to make: trust comes from the audit trail. She said it most simply with “A simulation has biases you can inspect. A human adviser has biases they can't.”

Some agreed, others pushed back.

We also had our partner, Mastercard, in the room with us. One of their senior leaders reminded everyone that trust at scale takes infrastructure. Card networks grew despite the distrust of plastic. They solved for trust through building rules behind the plastic - who pays when things go wrong, how a real merchant is identified from a fake one, and layers of security that assure counterparty transactions.  Decision AI in wealth has not yet built the trust layers and the perception of certainty. It will need similar scaffolding, and nobody yet knows who will build it.

Then a father in the room talked about his children. Their futures depend on choices he hasn't put on paper - when he plans to retire, what he intends to leave them, and when he intends to leave it. A wealth manager would draw it out of him over years of cups of tea. He wasn't sure a machine could, or should. This part felt similar to our conversations with retail technology leaders who sell, significantly, through in-store. They’re putting AI as close as they can to the customer, without them directly seeing it. It’s about maintaining the human connection in this, but empowering the client-facing staff with smart and considered insights at every interaction.

It would have been lovely if these different strands all fitted together neatly. But it’s never that easy. The scientist is right that decision AI is more auditable than the alternative. The card-network instinct is right that an audit trail without rules isn't trustworthy. The father is right that some of what humans know about each other never gets written down.

Which gives our ‘second-wave’ argument a sharper edge than we expected. The models are closer to ready than the trade press suggests. The harder, slower, more interesting question is who is best placed to build the trust scaffold for them, and on whose terms. And how will regulation support, accelerate or drag it.

Card networks took decades to build theirs. Wealth doesn't have decades if it is going to be ready for the great intergenerational wealth transfer. But it does have something the card networks perhaps didn’t have at the start - a regulator already in the room, advisers with a real point of view, and clients who are already telling you which conversations they'd rather have with a human and which they'd rather have with a machine. That's a head start. It's just not the same head start that most of the AI commentary is selling.

The firms that win this won't be the ones with the cleverest models. They'll be the ones who decide early what they want trust to look like, and then build everything else around that decision.

That, as it happens, is the conversation most boards are not yet having.