In Wealth Management, Trust Is Becoming Infrastructure

The Question I Left Open Last Month

I spent last month’s column on a version of this question in asset management, and I closed it with a line borrowed from McKinsey that I have not been able to shake: planning is getting cheaper than supervision. The value, I argued, was not disappearing. It was migrating towards the firms that own the data, the relationships, and the judgment that no agent can replicate on its own. McKinsey calls those defensible positions the control points.

That was the right frame for asset management. And the recent Hazel sell-off forced the same question into the open for wealth management, where the relationship is not a feature of the product—the relationship is the product. So what are control points, and how will they separate the winners from the used-to-bes?

Tasks Are Not Jobs

Let’s start with what the sell-off did not mean. It did not mean the advisor is finished. Rather it highlights how AI is exceptionally good at compressing the invisible labor of the industry–the tasks of preparation, data extraction, drafting, scenario modeling. Replacing tasks is not the same as replacing the human who is accountable for the judgment, the trust, and the behavioral coaching that clients actually hire.

The numbers support the distinction. McKinsey reports roughly 80% of affluent households still say they want a human relationship, and interest in holistic advice—the integrated, life-spanning kind that is hardest to automate—climbed from 29% in 2018 to 52% in 2023. Fee rates on relationships above $1 million have held remarkably steady at about 104 basis points since 2019. And the industry faces a projected shortfall of 90,000 to 110,000 advisors by 2034, which makes AI a capacity lever the industry needs, not a threat it can wish away.

So the replacement of humans question is the wrong question. The right question is as AI commoditizes the technical output, which parts of the value chain become mere features, and which become control points?

What Is A Control Point

A control point is a position in the workflow that stays scarce even after the surrounding work becomes free. The four below adapt McKinsey’s framework.

The first is governed data. AI is only as good as the permissioned, high-fidelity information it runs on. The firm that owns the proprietary client record—the full household balance sheet, the history, the context—owns the floor for every downstream tool.

The second is the interface. This is the advisor’s workspace, the screen where the relationship is actually managed. McKinsey’s sharpest warning is here: if an AI-driven operating system becomes the primary interface and a third party owns it, the incumbent’s century-old back office quietly degrades into a background utility. Hazel is interesting precisely because a tax tool is a plausible Trojan horse for the interface.

The third is auditability. For years the industry assumed regulation was a moat—that compliance was too hard to automate, so trust-based advice was safe. AI is dismantling that assumption by industrializing the oversight itself. The moat does not vanish; it moves to whoever builds auditability by design, i.e. evidence trails, supervision workflows, explainability that a regulator and a client can both follow. Compliance staff are a cost. Auditable systems are a control point.

The fourth is execution rails—the legal and technical ability to turn a recommendation into a completed transaction, securely. Insight is becoming abundant. The toll booth is the ability to act on it.

Underneath these four is accountable judgment–the named human who owns the decision when it matters and the behavioral coaching that keeps a client from selling at the bottom.

Notice what every item on the list has in common? Trust used to be a feeling—the advisor’s rapport and the brand’s reputation. In the AI era, trust is becoming infrastructure. Permissioned data governance, a defended interface, auditability, supervised execution—these are trust rendered as an operating system, engineered into the stack rather than carried in a relationship. The firms that win will not be the ones with the warmest handshakes. They will be the ones that have built trust into something they own.

What That Looks Like In Practice

This is no longer theoretical. JPMorgan’s asset and wealth management arm has been building exactly this, and its 2025 shareholder letter is unusually specific about it.

Its Connect Coach platform, launched just over a year ago, now serves roughly 12,000 users across the Private Bank and U.S. Wealth Management and runs 25 specialized AI agents that push around a million personalized, AI-driven insights to front-office staff. Advisors use it to accelerate meeting prep, support portfolio analysis, and generate call summaries—precisely the document-heavy tasks AI is expected to absorb first.

But the strategic point is not the productivity. It is who owns the layer. By building Connect Coach in-house, JPMorgan keeps the client data, the workspace, and the supervision under its own roof rather than depending on an outside vendor for them. That is an incumbent defending the interface control point: rather than let an independent platform like Altruist’s Hazel become the operating system its advisors work inside—and slide between the firm and its clients—JPMorgan owns that layer itself. The supporting numbers tell the same story: The firm’s private bank advisor headcount has nearly doubled since 2010 to about 4,100, while revenue per banker rose 15% from 2020 to 2025. Capacity is going up and the relationship is staying close. That is the disruption scenario—reshaped economics, human core intact—rather than the displacement one.

The Honest Counter

A skeptic would push back, and should. Control points are not permanent. McKinsey notes that baseline capability is inflating fast, so today’s differentiator—slick AI summaries, sophisticated tax modeling—becomes tomorrow’s table stakes. Fee pressure is real too, though likelier to surface as demands for transparency and unbundling than as a cut to the steady 104 basis points. And the incumbent’s edge is a path, not a guarantee: many firms still run on fragmented, decades-old systems. So the deeper danger isn’t that JPMorgan moves—it’s that most firms assume their relationships keep them safe and never build the infrastructure at all, which is McKinsey’s bluntest point: most are not yet building seriously for either scenario.

There is also a longer-horizon caution. As McKinsey notes in its 2035 outlook, the next generation of clients is shifting the basis of trust from “trust me” to “show me”—from reputation to demonstrated, auditable outcomes. That cuts both ways. It rewards the firms building trust as infrastructure, and it punishes the ones still trading on a logo.

The Race Is To Become Unrentable

The wealth management industry long assumed it sat safely on the relationship side of the automation line. This February the market called that assumption into question and priced in the doubt. The signal in the sell-off was not that advisors are obsolete. It was that the economics of advice are being rebuilt around what cannot be cheaply copied.

Producing the plan is getting cheaper by the month. Being accountable for it is not. Trust is no longer just something an advisor earns across a table—it is something a firm builds, owns, and is increasingly the only thing worth paying for.