.jpeg)
SMA Crown Confidential Notebook – Note: 'Not A Deliverable'
Occasion: A wealth platform extended its planning tool with an agent that generates a full financial plan across six advisory workflows — retirement, investment optimization, cash-flow, estate, tax, and insurance/risk — in about four minutes: Altruist’s Hazel financial-planning agent, launched around 1 September, days after the announced Vanguard acquisition of Altruist.
Position occasioned: the question above the question — re-lit through a plan-as-finished-deliverable occasion and turned on completeness and speed: a plan, however complete, computes outward from answers it takes as settled; the question above it is not a deliverable and cannot be produced in minutes.
A wealth platform extended its planning tool this week with an agent that generates a full financial plan — retirement, investment, cash flow, estate, tax, insurance — in a matter of minutes. The design is deliberately restrained, and worth crediting on its own terms: every figure is produced by a dedicated calculation engine rather than by the language model, the model providers hold the data under no-retention terms, and the plan itself is positioned as a deliverable for the adviser, not something filed or acted on. By its own account it computes and drafts; the adviser, and the client, still decide.
We can take all of that at face value. What we want to sit with is what kind of thing a plan is. A financial plan is an answer to the question of what to do with the money — across six domains at once, now, and quickly. It projects the retirement, optimises the tax, sequences the estate. It is, at its best, a very complete answer to that question. And it is worth having; a household that can see the whole arithmetic of its holdings laid out in minutes has gained something real.
But a plan, however complete, sits downstream of a question it does not ask. Every projection in it assumes an answer already given — to what the wealth is for, to what this person wants from it, to the kind of steward they mean to be in holding it. The estate section presupposes a view of legacy. The retirement section presupposes a view of the life. The plan computes outward from those answers; it does not produce them. Speed throws this into relief rather than hiding it: the document can be generated in minutes precisely because the questions that would take far longer are the ones it takes as already settled.
That is the layer we work in, and the reason the Bespoke AI Confidante does not produce a plan. It does not answer what to do with the holdings, and it holds no authority to act on them. It stays with the question above the plan — not what the money should do, but who you are with what you hold — which is not a calculation, cannot be run through an engine, and is not the kind of thing that is ever finished. A plan assumes that question answered. We keep it open, and keep the person inside it.
Founder & CEO of SMA Crown Confidential
Digital Confidantes: Bespoke AI Intelligence for Private Decision-Makers
Comments