Anthropic's advisor suite hands the AI edge to firms with clean permissions

Anthropic put Claude for Financial Advisors on the market Monday with nine launch partners — BlackRock, Charles Schwab, Addepar, Envestnet, iCapital, Orion, Wealthbox, Wealth.com and Zocks — and the roster matters more than the model inside it, because those firms hold custody, portfolio accounting, alternatives data, CRM and document workflow for many independent practices. An assistant that reads across all of them in one conversation is not competing with the advisor's stack so much as sitting on top of it.

The suite covers research, meeting preparation, portfolio analysis, documentation and compliance checks, and the announced capabilities are specific enough to test in a live book: pulling alternatives data from iCapital or Addepar, surfacing drift and concentrated positions against a client's target allocation, running estate and tax brief checks, and flagging content that may warrant additional compliance review. Access currently runs through Enterprise plans for RIAs.

The mechanism is worth understanding before the budget conversation. Connectors let Claude reach custodians, asset managers and wealth technology providers, and the plugin bundles advisor skills and those connectors into a single install, so the assistant reads systems the practice already pays for rather than a new data set. That trims the cost of trying it and raises the cost of having let those systems drift, since an answer assembled from a custodian feed and a CRM record is only as current as the least-maintained of the two.

Orion is where the picture gets concrete. The firm expanded its work with Anthropic and shipped a model context protocol connector into the plugin, so an adviser can ask for a client update, meeting prep or a read on the pipeline in plain language and get an answer assembled from live Orion Connect portfolio data and Redtail CRM records. The work it removes is the unglamorous kind — a report out of one system, relationship history out of another, context reassembled by hand. The connector respects existing Orion and Redtail permissions and gives firms firmwide control over access, which means an adviser sees what the entitlements already say they can see. Orion serviced $6.6 trillion in assets under administration and $211 billion of wealth management assets as of June 30, across more than 8.6 million technology accounts, and the connector is available now in Anthropic's connector directory.

That inherited permission model is the practical catch, and it is where a firm should start, because an assistant that reads through existing entitlements inherits the firm's access hygiene as it stands, including whatever stale permissions, broad view rights or outlived logins the practice has accumulated. Turning the connector on amounts to an access review most practices have been deferring, and it will surface what the firm's own records already contain, which is not a project any partner in this launch can finish on the firm's behalf.

BlackRock is moving onto the same surface, launching a plugin for its Advisor Center that joins plugins from S&P Global and LSEG that already sit alongside the Claude connectors. That puts asset managers in the position of competing for the first call an adviser's assistant makes, a different contest from shelf space on a platform; integration is the competitive ground for platform providers, and a position bought early inside a default is hard to unwind later.

Conquest Planning went the other direction, giving independent advisers and RIAs direct access to its full platform, including its Strategic Advice Manager planning engine, designed for firms with 10 or fewer advisers and delivered through digital purchase, setup and onboarding, with launch integration partners including Jump, Morningstar and Schwab Advisor Services. Small practices get a planning engine they can switch on without an enterprise contract, while larger ones get an assistant wired into custody and portfolio accounting. Both are chasing the same hours, the ones between the last meeting of the day and the first one tomorrow.

Meeting prep is on Anthropic's list of uses, and one number should govern how a practice deploys it: Jump, a launch partner on the Conquest side, analyzed nearly 12,000 client meetings and found advisers talking more than clients in 84% of them. A better brief does not change who does the talking, so a practice that installs the feature without deciding who owns the meeting will get the same transcript it always got.

Two bundled skills carry more weight than the rest. The estate and tax brief check arrives in the same conversation as the portfolio review, and as this publication has argued, tax planning is the last high-margin skill in a portfolio business that has commoditized everything around it; the brief itself is the cheap half of that work, while sequencing, exemption decisions and the conversation about what the numbers mean for one family are where the fee survives, and a model that drafts the brief faster leaves more room for them.

Compliance review is the other one, and it is where this publication's standing view on AI adoption applies most directly: the suite can flag content that may warrant additional compliance review, but the supervision, the sign-off and the record of who reviewed what stay with the firm. Firms that buy the bot before building the sign-off pay twice, and the second bill tends to arrive as an exception someone has to explain, long after the efficiency gain has been booked.

Anthropic's own advisor-facing arrangements are still being priced, and this publication has described a process in which wealth managers submit pricing and service proposals for the firm's advisor bench, arguing that it will be won on fees, not logos. The same commercial question now sits under the distribution list: nine connectors plus the plugins from BlackRock, S&P Global and LSEG are arrangements as much as features.

What the suite costs is not in the announcement. Access through those plans is a gate as much as a price, and any practice that has absorbed usage-based AI language into a platform renewal knows the shape of what follows: as this publication has argued, the meter turns whether or not anyone uses the feature. The connector directory is live and Orion's sits in it, so an Orion shop can now hand the model a pipeline review assembled from its own CRM records; whether that answer is worth anything is a records question before it is a model question, and the state of those records is the one variable Anthropic did not put in the directory.