Schwab Advisor Services becomes first RIA custodian to use Anthropic's Claude
Schwab says it is the only RIA custodian in Anthropic's early release, though Vanguard is on the charter list and says it will not use Claude for custody.
Schwab Advisor Services has become the first RIA custodian to use Anthropic's Claude for Financial Advisors, and its release identifies the firm as the first and only RIA custodian included in the early version, a distinction RIABiz, which reported the launch, frames as exclusive for now. Anthropic announced Claude for Financial Advisors alongside a charter list that includes Schwab and Vanguard Group, and Peter Nolan, the vendor's head of asset and wealth management, described the collaboration in terms of reach: "Collaborating with Schwab and the 16,000-plus independent RIAs it serves brings Claude to the firms doing that work every day, starting with a clear view of what has changed for each client."
Joel Bruckenstein, producer of the T3 conferences, reads the launch as distribution rather than product. "Schwab is in the driver's seat now," he said, according to RIABiz. "Schwab is the gorilla, and they're going to get first crack at taking it to 16,000 advisors." In his account the custodian ends up holding the tool RIAs will be obliged to adopt, a claim about market pressure rather than any contractual term. RIABiz reports that other analysts counter that the launch is generic and late-coming.
The two readings can coexist. A tool that looks ordinary on its own can still matter when it arrives as the default inside software a firm already touches every day, and first access to 16,000 advisors is a distribution question rather than a model question. Bruckenstein's further point, that RIAs will be compelled to adopt it or else, suggests a bundled assistant becomes one more line on a custodian's recruiting sheet. Late-coming is a fair verdict if the comparison set is other AI vendors, and generic is a fair one if the standard is capability rather than placement; neither critique reaches the question of who holds the distribution, which is the part Bruckenstein is pricing.
The license is the cheap part
For an advisor, what arrives is a platform feature rather than a purchase decision, which makes the seat price the least interesting number attached to it. As this publication reported when the integration surfaced, the arrangement turns AI into a custody line item, and the process work around it costs more than the license does; our earlier reporting put the seat at $240 a year.
The coverage does not itemize that process work, and it is the part a practice has to fund. RIABiz notes the launch landed amid wider concern over whether humans keep control of AI systems; inside a firm that concern takes the shape of decisions about what client information the model may see, who reviews its output before a client does, and how any of it reaches the supervisory file. The burden falls on whoever clicks approve, and buying the tool through the custodian does not move it off the advisor's desk.
Adoption will not be uniform, either. The split runs between practices that convert a new tool into leverage and practices that buy it and change nothing, and the line is drawn not by how good the model is but by whether the firm rewrites the routines the tool touches; a seat that gets billed and never opened is the purest form of the second group. The routines are unglamorous and they are where the money is. A firm that defines them once, deciding what client data goes in, who reads the output and what gets logged, collects the benefit of every model improvement the vendor ships afterward, while a firm that leaves the question to individual advisors gets inconsistent answers and a supervisory file that will not hold up on examination.
What Vanguard put into Claude
Vanguard sits on the charter list too, which complicates Schwab's exclusivity on its face and not much beneath it. Vanguard said it will not use Claude for RIA custody or for its in-house advisors, and pointed questions toward Anthropic, whose release describes Vanguard's contribution as bringing model portfolio and asset allocation information into the tool so advisors can draw on it when making client decisions. That is a research arrangement, not a custody integration, which suggests a custody-scale deployment is what would make Claude an operating standard rather than a research add-on.
Which Vanguard advisors might touch it is unresolved. RIABiz raises two candidates, advisors inside Vanguard Personal Advisor Services or third-party advisors on Altruist, the RIA custodian Vanguard is acquiring, and the coverage does not say which.
RIABiz also reported that Fidelity declined to respond to a query about where its RIA custodian stands with AI or Claude, and the coverage does not say where that leaves Fidelity's advisors. Fidelity has been making its own platform decisions in public view: as this publication has reported, the firm told custody clients under $100 million they have until June 2027 to move, with no fee option to stay. Schwab is adding capability to a base of 16,000-plus RIAs, while Fidelity has been narrowing who sits in its base.
That divergence is the part a practice can plan around. Shelf access, sweep revenue, referral arrangements and now AI defaults are all terms on which platforms compete, and custody has stopped being neutral infrastructure. An assistant installed as a platform default is the first of those terms that changes how the work gets done rather than what it costs.
The calls an advisor can make now are narrower than the analyst debate: whether the seat sits inside the platform fee or on its own line, who reviews what the model produces before a client sees it, and which routines change once the tool becomes the path of least resistance. Whether the early-release price survives the end of the window is left open.
Save this analysis and keep the funds you follow together in My Desk.
Sign in to save articles or follow funds.