.jpeg)
Who Sets the Dial
The same AI engine acts for the bank and prepares for your adviser. On your side of the table, the dial should be yours to set.
On 14 September 2026, Anthropic released Claude for Financial Advisors. It gives an advisory firm eight ready-made skills, among them pre-meeting preparation, post-meeting notes and follow-up, an estate and tax brief and a portfolio rebalance review. It also adds eleven connectors to the systems advisers already use, including Charles Schwab, BlackRock, Vanguard, Addepar, Envestnet and Orion. Through the Schwab connector the model can see a client’s balances, positions, transactions, cost basis and the status of money movements. One sentence in the release sets the limit: investment recommendations, client communications, compliance determinations and other regulated activities “remain subject to human review and approval.”
Seven months earlier, the same company’s model appeared in a different setting. In February 2026, Goldman Sachs was reported to be building autonomous agents on it for trade reconciliation, accounting and client onboarding, and the bank’s chief information officer called them “digital co-workers.” The reports describe agents that carry out multi-step processes. For the adviser, the engine prepares and stops. For the bank, it acts.
As such, none of this is extraordinary. A provider selling different arrangements to differently regulated buyers is ordinary commerce, and the limit drawn for advisers is a sound one. What the two announcements show together is that the limit is not a property of the engine. It is a setting, a dial, and it is turned to suit whoever is buying.
Now take the adviser’s product from the chair of the principal — your chair. You are not its user. You are its content. The balances, positions and transactions flowing through the connector are yours. The “pre-meeting prep” skill is preparation about you, and the “estate and tax brief” is a brief on your estate. Your adviser now walks into the room having been briefed, in minutes, by a system that has read your accounts and the notes of your last conversation. One of the connectors is Zocks, the meeting-notes tool from the first piece in this series: the minutes feed the preparation.
Then look at the dial. For the bank it is set to act, because the bank answers for its own back office. For the adviser it is set to prepare and stop, because the adviser answers to a regulator for what reaches a client. Each setting follows the liability of the party that bought it. The principal has never been a buyer in this market, so no one has set a dial for you. No one has asked what the engine should prepare on your side, what it should question, and where it should stop and hand the decision back.
Here the news turns in your favour. The engine is not the scarce thing. Advisers do not receive it from their custodian. They subscribe to it directly, as anyone can, and the same model may already be on your laptop. What the adviser receives in addition is the arrangement: skills written for an adviser’s work, and data wired to an adviser’s desk. A product leader at Vanguard made the point from the other direction, warning that “slapping a generalized LLM on top of legacy infrastructure simply doesn’t solve the core issue.” The point holds on your side of the table too. Access is not capability. An engine with nothing arranged around it answers whatever it is asked and prepares no one.
From your chair, the principal’s, the questions are these. Which AI prepared your adviser for your last meeting, and what did it read about you? What was prepared on your side? If the same engine sits on your own laptop, what has it been formed to do for you? And who decided where it stops?
The first answer is practical. Ask your adviser which AI tools now touch your data, what they prepare, and which acts still require a person’s approval. A firm that has drawn its limits with care will be able to tell you.
The deeper answer is to set a dial of your own. The same engine can be arranged for the principal as deliberately as it has been arranged for the adviser: formed around your affairs and intentions, ready with your questions before a meeting, reading what was agreed after it, and stopping where the judgement is yours, which is at every decision. That arrangement does not come in the box. It has to be authored for one person, and tended as circumstances change.
That is what SMA Crown Confidential forms. Not another adviser at the table, but an intelligence layer on your side of it: a Bespoke AI Confidante running on the model of your choice and formed around your own intentions, so that your judgement stays independent of every firm that has arranged the same engine for itself. The engine is the same on both sides of the table. The question is who sets the dial on yours.
Founder & CEO of SMA Crown Confidential
Bespoke AI Intelligence for Private Decision-Makers
Sources: WealthTech Strategy, “Claude for Financial Advisors,” 17 September 2026; RIABiz, 15 September 2026; CNBC and MLQ on Goldman Sachs and Anthropic, 6–7 February 2026.
Comments