The AI sell-off reprices the cash spread, not the fee

LPL Financial and Charles Schwab each fell more than 6% on Tuesday as investors worked through reports about Meta's Muse, the consumer AI application whose swift adoption has changed what public markets expect from any business that charges for human judgment, while the S&P 500's financials sector dropped 2% over the same session, InvestmentNews reported. It is the second time this year that a product launch has knocked the shares of publicly traded advice firms, following the selloff that came when fintech Altruist unveiled a tool aimed at personalized tax strategies, as the Wall Street Journal and others noted.

The stated fear is easy to name and hard to price: advisors bill clients in the neighborhood of 1% of assets annually, and the worry is that AI-driven financial planning erodes both that fee and the earnings of companies that lean on advisors today for commissions and fees—think LPL and Ameriprise among the publicly traded set. That fear made Tuesday's selling broad enough to catch advisory firms and custodians in the same net and to drag the whole financials sector along with them. But Devin Ryan, who heads financial services and fintech research at Citizens, told the Journal that no single AI agent is upending the sector overnight, but that a run of product launches has "allowed a narrative to run that the world is changing and there's more uncertainty as the world changes." In Ryan's view, agents could eventually move client money for tax-loss harvesting and other goals efficiently enough to leave less cash sitting around for brokerages to turn into their own profits. Tuesday's sellers were pricing a fee problem, but the mechanism Ryan describes is a float problem, and it likely bites sooner.

A firm with LPL's balance sheet does not need its planning fee to compress before agents show up in the income statement; it needs only for the cash clients leave parked on the platform to move into something that pays the client more and the firm less. LPL carries $819.1 billion in registered assets across roughly 2.85 million accounts, per PWD's records, and the yield on whatever slice of that sits idle is a revenue line the fee debate never touches. The exposure is old news; the new piece is that an agent can act on that cash without the client ever changing advisors.

The launches behind the narrative keep arriving in the channel where RIAs actually live. Last week Anthropic rolled out a suite of Claude workflow tools that connect to platforms run by Schwab and Vanguard; last month Vanguard said it would buy Altruist for $4 billion; and Muse has already partnered with PayPal and Plaid, according to the Journal's reporting, with analysts there seeing it reaching basic consumer payments, planning, and insurance coverage.

As this publication argued when Schwab took its Claude arrangement, the custodian bought queue position, not a better model, and queue position is worth whatever the plumbing beneath it is worth. The price Vanguard agreed to pay for Altruist makes the related point in dollars: the client record is the asset no model partnership can purchase, which is why a billion-dollar RIA leaving Schwab for Altruist's platform, as PWD reported, carries more weight than any product demo.

LPL has been answering on the talent side. In August it hired Wells Fargo's technology chief, Jonathan Lewis, as chief technology officer, betting that platform engineering becomes the next retention weapon—a reasonable bet pointed at a different target, since software that keeps advisors in their seats does not defend the spread on cash that clients decide to move, and only one of those two problems surfaced in Tuesday's quotes.

The tension in the tape is hard to miss: the same investors who took 6% off the incumbents have been rewarding the platforms that assume software carries more of the work, and both reactions can be right at once, since a fee that compresses for the incumbent is share that expands for whoever owns the connector. What they cannot both be is equally urgent next quarter.

The sell-off is right in direction and imprecise in what it names: something real is being repriced in the economics of holding client assets, and firms earning a spread on parked cash will feel it before firms earning a planning fee. The 1% fee, though, is the hardest part of this model to automate, because it pays for judgment, for the relationship, and for the meeting where a client signs off. That last piece is where this publication expects the next product fight: the confirmation meeting, where an AI-drafted plan meets a human signature and the fee gets justified all over again.

Watch LPL's next disclosure, and specifically how much client cash is still parked on the platform. If that number holds through the quarter, the market marked down the wrong line.