When the wealth manager opened the tool for the first time, the client’s assets were already there. Balances, holdings, and transaction history, pulled straight from Charles Schwab, sat on the screen next to a set of talking points drafted by Anthropic’s model. Nothing could move, though, until a human adviser said so. That arrangement sits at the center of Claude for Financial Advisors, the product Anthropic began rolling out to wealth management firms on September 14.
The system is built for the preparation and paperwork that fill an adviser’s day. It assembles material for client meetings, pulls together research and administrative documents, and drafts the messages that go out to clients. Schwab acts as the custody partner, wiring the tool into client account balances, positions, and money movements. BlackRock and Vanguard supply portfolio analysis and research. Addepar and iCapital provide the software platforms on which those pieces sit.
The guardrails are the telling detail. Before the system executes any action, a human adviser must explicitly authorize it. Every step of the workflow is logged. Copy intended for clients is automatically checked against the SEC’s marketing rules before it goes out the door. Anthropic is selling automation, but it is selling it with a human hand on the switch.
Peter Nolan, who leads Anthropic’s asset and wealth management group, described the release as “the infrastructure for AI-native wealth management practices.” The wording is deliberate. Anthropic is not pitching a chatbot. It is pitching rails, a set of connections through which a model can touch client money without touching it unsupervised.
The market Anthropic is chasing is not small. Wealth management firms in the United States administer trillions of dollars in assets, and their back offices run on a blend of spreadsheets, legacy systems, and manual review. A tool that shaves an hour off an adviser’s week compounds quickly across a firm with thousands of advisers. That arithmetic, more than any benchmark, is what Anthropic is selling.
Anthropic has been building toward this kind of product for months. The company has pushed Claude into the enterprise through a series of integrations, positioning the model as a workhorse for corporations rather than a consumer curiosity. Financial services is the natural next step, because the industry pairs high document volume with strict compliance demands, a combination that rewards a model trained to be careful.
The launch puts Anthropic on a direct course with OpenAI, which introduced a finance-industry tool of its own only a week earlier. The two labs have spent years competing over who can train the smartest model. That contest is now leaking into the quieter business of enterprise software, where deals are won on integration and compliance checklists rather than benchmark scores.
The rivalry adds urgency. Both companies are chasing the same pool of enterprise budget, and both have discovered that a general model needs a specific wrapper to clear a bank’s procurement process. Whoever signs the most custodians and research providers first may set the pattern that the rest of the industry follows.
Anthropic’s choice of partners says something about how it expects to win. Schwab is one of the largest custodians of retail wealth in the United States. BlackRock and Vanguard sit on the other side of the table, supplying the research that advisers lean on to justify their decisions. By plugging into all of them at once, Anthropic positions Claude not as a replacement for the adviser but as a layer between the adviser and everything the adviser already uses.
The timing is not accidental. Wealth management has become a testing ground for AI because the work is document-heavy and the underlying data is already structured. Balances, positions, and trades live in standardized formats that a model can read without much coaxing. The harder problem has always been trust. A model that drafts a note is harmless. A model that moves money is not. Anthropic’s answer is to draw the line at the adviser’s approval and to make that approval traceable.
Analysts said the approach reflects a wider shift among AI labs toward industries where regulation is thick and the stakes are high. Selling to banks means accepting audits, data-retention rules, and the slow rhythm of enterprise procurement. In exchange, the labs get contracts that recur for years. The alternative, selling consumer subscriptions, is faster but far less sticky.
The wealth business has reason to be cautious. It has watched other corners of finance adopt technology quickly and then spend years cleaning up the mess. Advisers, for their part, are protected by a simple fact: clients want a person, not a model, to answer for their money. That is why Anthropic is leaning into the adviser’s judgment rather than trying to route around it.
Whether advisers take to the tool is the open question. It does not change what an adviser does. It changes how fast the adviser can do it. For a firm trying to serve more clients without hiring more people, that is the point. For a client wondering whether the human is still making the calls, the answer, at least for now, is yes.
