Expertise at a fraction of the cost

Family offices, as Michael Macfarlane observes, tend to remain small, trust-based teams relying on a handful of advisers. Lawyers, accountants and other specialists supply knowledge the family does not hold itself, and the relationship rests on trust. Three Wealth Office conversations, with Shield Xiao Dun, Nicolas de Santis, and Lord Bernard Hogan-Howe with David Ferbrache OBE, approach artificial intelligence from very different directions. Read side by side, they say something specific about that circle.

Xiao Dun's subject is education, but his argument concerns expertise in general. His first company, 17 EdTech, grew to 20,000 staff. His new company in Palo Alto builds AI tutors that cost a fraction of hiring the best human tutor, and he describes the effect as something like having Einstein teach you physics or Chopin teach you the piano. When Macfarlane observes that AI now does what 20,000 people did less than a decade ago, Xiao Dun agrees, and suggests that anyone can now have a team of that size to build or create with. He points to the heavy use of general-purpose chatbots by adults for learning, writing help and question-and-answer.

De Santis, who co-founded Opodo and helped launch one of the earliest digital currencies, offers a cooler reading of the same technology. AI is not new, he points out. It has been part of computer science courses for around 30 years, and the current boom is a bubble, as booms in many industries are. His own advisory work with families that own large corporations rests on something he says is taught in neither schools nor business schools, the ability to imagine and build a future beyond the family's past. Neither guest suggests that knowledge has stopped mattering. Their accounts imply that access to it is becoming cheaper, while judgement about what to do with it remains scarce.

Judging advisers by what they do

If expertise becomes more widely available, the question of how to judge the people who supply it grows in importance. Xiao Dun approaches it from hiring. Interviews, he says, are among the more modern ways to assess people, yet still not very effective.

I mean, I've interviewed lots of people and made mistakes many, many times.

— Shield Xiao Dun, Wealth Office

He finds it far more reliable to judge someone by their output over time, in the manner of a work trial. His interest is in how schools could assess students once AI makes that kind of evaluation practical.

Hogan-Howe reaches a similar conclusion about security advisers by a different route. The worst consultancy he has seen hands over a long list of tasks and leaves the client with the same problem and more to worry about. He prefers an initial assessment of gaps followed by advisers who work alongside the team, where a client can answer a recommendation with an alternative that suits it. He also stresses timing. A family that recruits help on the day of an incident may find it cannot build a trusting relationship quickly enough. Both men, in other words, favour evidence gathered over time from working together, and distrust judgements made in a single encounter.

The adviser as a route in

The cyber conversation adds a harder edge. Ferbrache describes attackers who compromise law firms and accountancy firms so that they can email clients with all the credibility those firms carry.

So we're seeing that mix of professional services also being targeted, because sometimes those advisors are actually more vulnerable than the final organisation as well, but they're trusted.

— David Ferbrache OBE, Wealth Office

AI, in his account, makes this easier. It helps criminals research targets, construct credible personas and produce well-crafted phishing lures, sometimes with simulated video or audio. Hogan-Howe adds that AI is starting to defeat biometric safeguards by replicating a person's voice and appearance. He also notes that the trust within a small office can amplify an insider threat, because people come to trust each other so much that they take risks they should not.

The implication for the adviser relationship is uncomfortable. The trusted channel through which advice arrives is itself a target, and a convincing voice or message no longer proves who sent it. Hogan-Howe's response is to test arrangements before they are needed. A principal, he suggests, can ask exactly who will be contacted, and which organisation will provide support, if an incident strikes abroad. If the answer sounds woolly, more work is required.

What remains with the family

The three conversations converge on what the family cannot hand over. Ferbrache argues that cyber security cannot be delegated to IT alone, and that because reputational damage falls on the family, the principal needs to be involved in thinking through the worst-case scenarios. Macfarlane adds that a principal may find it hard to disclose personal vulnerabilities to people they work with every day, and Ferbrache agrees the engagement should often sit with the principal directly.

Xiao Dun locates the irreducible part in purpose. The two dystopian futures he fears from AI, one of passive consumption and one of extreme concentration of power, might in his view be more profitable than the path he prefers. He holds that humanity is a larger concept than technology, and that products should serve it. De Santis, asked what a young member of a wealthy family should do, answers that they should get their hands dirty and focus on value creation.

None of the guests describes advisers becoming unnecessary. Their combined account points somewhere narrower. AI lowers the cost of knowledge and raises the cost of misplaced trust. Families will still need advisers. On this reading, more of the work falls to the family itself, in judging advice by its results, verifying the channel it arrives through and keeping decisions about purpose and exposure in its own hands.