AI has entered the client meeting, and fees haven't moved

The plan is already written before the advisor speaks, because Altruist's Hazel and april can pull IRS transcripts by wire and the planning agents now draft in real time while the client sits across the table. The advisor's job becomes deciding what happens after the machine has run, a pricing signal the industry has not yet paired with a fee conversation. For a decade, the RIA's core deliverable was the plan — the tax projection, the estate flow, the asset location — and once a model produces the plan in the meeting, that deliverable becomes a commodity, while the fee has to be defended elsewhere.

SEI's survey of 518 advisors and 302 wealthy investors found clients want the tax figure first, and roughly half of advisors cannot produce one — a workflow gap. The planning agent can produce the number in the meeting, but the advisor still has to verify it against the client's actual filing, explain the trade-off, and document the advice. The client pays for the confidence that the number is right and the tax is actually saved, rather than the number itself, and the demand is for a specific figure while the supply of advisors who can deliver it on demand remains half the market.

Mercer's survey found 63% of asset managers are running off-the-shelf AI tools, a number that should reframe how allocators evaluate managers and how advisors evaluate their own tech stack. When a majority of managers share the same underlying models, the capability becomes a correlation trade rather than a differentiator, and the managers who charge a premium for proprietary AI are selling something the survey says most of their peers have already bought. Fees have not moved, because the AI stack has been priced as a vendor line item instead of a change to the deliverable. The next repricing will come from the client question about what the fee buys, independent of the software contract.

Vanilla, SS&C and FinTurk are pushing agents across the whole book as the budget share dedicated to AI has doubled while firms wait for proof that the tools create revenue rather than just slide work around. The RIAs that buy the detection without the capacity to clear what it finds will pay for it twice: once for the software, and again for the liability or the client churn when the flagged issue sits unaddressed. The hidden tax of the AI buildout is that the tool finds a problem, the firm cannot staff the fix, and the client discovers the gap in the next annual review; the whole-book agent widens it instead of solving it.

Luminary's $22 million round puts Rockefeller, BNY and Focus on the cap table of the software that decides where a client's assets go at death — the estate-planning agent moving into the meeting, and the same trade from the other side: the platforms are buying the file rather than the advice. The advisor who sees this as a threat misses the point; the file is cheap, the advice is expensive.

Hamachi's patent is the compliance choke point that makes that liability explicit: the award covers one method of masking client data before a model reads it, narrower than the sales pitch implies, and the narrowness is the entire point. A patent on one masking method means the vendor can protect a specific way of keeping data inside the RIA's boundary while the model runs elsewhere. The right question for any advisor signing an AI contract is where the data sits while the model runs, more than what the model can do. If the data leaves the RIA's environment, the RIA retains the liability while the vendor retains the model; if the data is masked in place, the RIA has bought itself an audit trail. The asset that commands the fee is the firm's ability to stand behind the output, more than the software itself.

Vanguard's $4.6 billion payment for Altruist's account rail sharpens the same point from the custody side: the RIA chiefs who welcomed the purchase priced it as a vendor story, but the referral economics on their own panel say it is a fee story. When the account rail itself becomes the product being acquired at that price, the margin available to the advisor for pure planning advice is the variable under pressure. Clients, especially younger ones, are already using AI to interrogate the fee, and a custodian survey found the youngest clients churning hardest while telling advisors the answer is digital. That behavior rewards the advisor who can defend the fee with a specific, tax-grounded plan and the operational record to match, more than the advisor with the best model.

The premium is shifting to the capacity to clear what the AI finds, a staffing and compliance decision. The 2027 budget line for agentic AI belongs behind the advisor, in the work around the meeting, rather than in the meeting itself. Three wealth-tech leaders put the return there, making next year's technology line item a staffing decision before it is a software one. The firms that understand this will stop pricing planning as a deliverable and start pricing the resolution of machine-flagged risk as the product. That is the trade the market has not yet made, and it is the one that will separate the RIAs that hold their fee from the ones that watch it leak into the custody rail.

The Hamachi patent matters because it forces that question into the contract. A model that can read a client's estate and flag a tax exposure is cheap once the same model is running at 63% of asset managers. A firm that can receive that flag, confirm it against the client's actual documents, execute the fix, and carry the professional liability for the advice commands a premium. The fee, in other words, is for the clean handoff between the AI's finding and the client's signed plan. The industry has spent two years buying the detection; the next twelve months will be spent discovering who can clear it. Watch the firms that are hiring estate attorneys and tax specialists ahead of the next model announcements.