Claude for Financial Advisors Isn’t Really About Finance

A client review is booked for 11 o’clock. The meeting itself will take forty-five minutes. Getting ready for it will take most of the morning.

You pull positions from the custodian. You check what moved since the last review. You look for the notes from the previous meeting, which are in the CRM, or in an email thread, or in a colleague’s head. After the call you write the summary, send the recap, log the follow-ups and hope compliance doesn’t object to one line in the email. Kitces research, which Anthropic quotes in its launch post, puts a number on it. A typical advisory practice spends only about a sixth of its time actually in client meetings.

If you work in marketing ops, you know this shape of day very well. The campaign idea takes twenty minutes. Getting it through the stack takes three days.

On 14 September, Anthropic launched Claude for Financial Advisors, a set of connectors and skills aimed squarely at that prep-and-paperwork problem. Most people will file it under “AI for wealth managers” and scroll past. I think that misses the interesting part. Take out the custodians and the SEC references, and what’s left is the clearest template so far for how AI will actually arrive in any job that runs on five systems and a compliance team. Marketing is one of those jobs.

This is an assessment based on Anthropic’s launch announcement and the partner statements published with it, not a hands-on review. I’m not a financial advisor, and nothing here is investment advice.

What Anthropic actually shipped

There are two pieces, and they matter for different reasons.

The first is a set of connectors. These let Claude read from, and in some cases act on, the systems an advisory firm already runs on. Charles Schwab brings in custodial data like balances, positions, transactions and cost basis. Orion brings portfolio reporting plus Redtail CRM. Envestnet covers Tamarac accounts with a MoneyGuide plan snapshot. Addepar, SS&C Black Diamond, iCapital (for alternatives), Wealth.com (estate and tax), Wealthbox (CRM), Zocks (meeting notes), BlackRock and Vanguard (model portfolios and research) make up the rest. They sit next to connectors Claude already had, including Microsoft 365, Salesforce, DocuSign, Box, FactSet, S&P Global and Morningstar.

The second piece is a set of eight skills, packaged with the connectors into one plugin you install inside Cowork. Each skill is tied to a specific moment in an advisor’s day: onboarding a new advisor, prospect intake, pre-meeting prep, post-meeting notes, a portfolio rebalance review, an estate and tax brief, an alternative investments brief, and a compliance and AI policy check. Firms can use them as they come or adapt them to their own workflows and house style.

On access, Anthropic recommends the Enterprise plan for registered investment advisers because it includes the audit logs that support recordkeeping. If you already have Enterprise, you find the plugin in the Cowork plugin browser and go through a guided setup to pick which tools to connect.

The chatbot is the least interesting part

Nobody in wealth management was waiting for a model that could write a meeting summary. Claude could do that two years ago. What it couldn’t do was see the Schwab positions, the Redtail notes, the estate plan in Wealth.com and the transcript from Zocks at the same time, for the same client, without someone copying and pasting between five tabs.

That’s the actual product here. The connectors are the moat, and the chat window is just where you happen to type.

Look at who signed up. Schwab, Envestnet, Orion and Addepar sit underneath a big slice of the US advice industry, and BlackRock and Vanguard are two of the largest names in asset management. Each of them is now choosing to put its data inside someone else’s interface. Some are going further and shipping their own plugins; BlackRock is launching one for its Advisor Center, next to existing plugins from S&P Global and LSEG. One advisory firm’s technology head said in the launch post that problems which once took months of configuration and outside partners can increasingly be solved by their own developers.

The timing is telling too. Within a fortnight of this launch, Anthropic announced a Claude Marketplace for partner plugins and agents, and a push to get more companies building plugins. Vertical bundles like this one are how that marketplace gets filled.

If you run a marketing stack, this is the part to sit with. The question is no longer “which AI tool should we buy”. It’s “which of our systems can the AI actually read, and with what permissions”. Salesforce, Microsoft 365 and Box are already connected. Your marketing automation platform, your consent tool and your CDP will each have to decide whether they want to be the thing the AI reads from, or the thing someone exports a CSV out of. The vendors that pick the second option will find the choice gets made for them.

Skills named after a calendar, not a feature list

Read the skill names again. Pre-meeting prep. Post-meeting notes. Prospect intake. They read like entries in an advisor’s diary, not like a product spec.

That’s a deliberate choice, and I think it’s the smartest thing in the launch. Most marketing AI is still sold as features: generate subject lines, summarise this report, write five variants of the ad. Useful, but you still have to work out where it fits in your day. These skills start from the moment (the hour before a client walks in) and work backwards to which systems need to be read to get it done.

Now put the same lens on marketing ops. Almost every advisor skill has a direct twin in our world:

Advisor skill	What it does for the advisor	The marketing ops twin
Pre-meeting prep	Pulls holdings, recent activity and open items into one brief	The account brief before an ABM or sales meeting: engagement, open opportunities, last conversation
Post-meeting notes	Turns a transcript into a summary, a recap email and CRM tasks	Post-demo or post-event follow-up, logged in the CRM with drafts ready for the SDR
Prospect intake	Organises what a prospect shares into a summary and a handoff memo	Lead handoff with enrichment and context, so the first call isn’t cold
Portfolio rebalance review	Flags drift from the target allocation and drafts the explanation	Budget and pacing review that flags channels drifting from plan and drafts the “why”
Estate and tax brief	Checks the estate plan against actual account titling and flags mismatches	Checking the CRM against the automation platform and consent tool, and flagging where they disagree
Compliance and AI policy	Screens client-facing language against the SEC Marketing Rule	Brand, legal and consent review before anything goes out
Advisor onboarding	Connects a new advisor’s tools and runs their first meeting prep	Getting a new marketer productive in week one, not month two

The row I’d pay attention to is the estate and tax one. It’s the least glamorous skill in the bundle, and probably the most valuable. It doesn’t write anything clever. It checks two systems against each other and tells a human where they don’t match.

Anyone who has run marketing automation for a few years knows that most of our worst days come from exactly that kind of mismatch. A contact opted out in one system and not the other. A lead status that means one thing in the CRM and something else in the automation platform. An AI that quietly catches those before a campaign goes out is worth more than one that writes a nicer email.

Anthropic built a marketing compliance reviewer and hid it in a finance product

Here’s the detail most marketers will skip. The compliance skill screens client-facing language against the SEC Marketing Rule.

That rule governs how US investment advisers advertise: testimonials, endorsements, performance claims, the lot. So one of the eight skills in a “finance” plugin is really a marketing compliance reviewer. It flags content that may need a closer look based on criteria the firm sets, helps document the review, and includes a workflow for writing down how the firm uses AI in the first place.

The rest of the design follows the same instinct. Claude prepares briefs and drafts. CRM updates and client emails are staged for the advisor to approve, not sent. Anthropic is explicit that recommendations, client communications and compliance calls stay with a human.

I like this design. I also said last week, while writing about Adobe CX Enterprise Coworker, that human in the loop is a setting, not a guarantee, and that applies here too. An approval screen is a checkpoint. It only becomes a control when the person approving actually reads. Put forty drafted follow-up emails in front of a tired advisor at 6pm and watch how fast “review” turns into “approve all”.

There’s a second catch in the phrase “criteria the firm sets”. The screen is only as good as the rules someone wrote down. A lot of compliance teams, in finance and in marketing, run on judgement that has never been written anywhere. An AI reviewer will expose that very quickly.

And note the plan requirement. Anthropic recommends Enterprise for advisory firms because that’s where the audit logs are. If your industry expects you to show who approved what and when, the cheaper plan isn’t really cheaper.

Hours, not headcount (and where I’d push back)

The best line in the launch post isn’t from Anthropic. It’s from Michael Batnick, managing partner at Ritholtz Wealth Management: “The math on this business is hours, not headcount.”

That’s the right way to pitch this, and it’s more honest than most AI launches. Nobody is claiming the advisor goes away. The claim is that an advisor has a fixed number of hours, most of them currently go to prep and paperwork, and some of those can come back. Anthropic pairs it with a demand-side number from EBRI’s 2026 Retirement Confidence Survey: more than four in ten American workers say they don’t know who to go to for good financial or retirement advice. So the argument is about capacity, not cost-cutting.

I buy the direction. I’d still push back on four things.

First, every voice in the launch has a stake in it. There are eighteen partner and customer quotes, and not one number on time actually saved. No pilot data, no before-and-after. That will come, but until it does, treat the quotes as intent, not evidence.

Second, a brief is only as good as what it’s built from. The pre-meeting prep skill pulls from the CRM, the custodian and old meeting notes. If the CRM notes are three months stale, the advisor walks in confidently wrong. That’s worse than walking in knowing you haven’t prepared.

Third, saved hours don’t automatically become client hours. A firm can just as easily use them to load more households onto each advisor. That might be good, more people getting advice, but it’s a business decision, and the software doesn’t make it for you.

Fourth, a single AI layer that can read custodial data, the CRM, estate documents and meeting transcripts is powerful and a serious access question. It’s not a coincidence that several partner quotes go out of their way to mention existing permissions and controls. If I were running IT at one of these firms, how each connector is scoped would be the first thing I’d check, well before any skill.

What about India?

As shipped, this is a US product. Every custodian and wealth platform in the connector list is American. There’s nothing for CAMS or KFintech, nothing for Indian brokers, nothing for the CRMs most Indian advisory practices actually use. A SEBI-registered investment adviser installing it today would get the skills and the general connectors like Microsoft 365 and Salesforce, and not much of the finance-specific plumbing that makes the product interesting.

The regulatory side is closer than people think, though. In 2025 SEBI amended its Intermediaries Regulations to add a chapter on the use of AI. Regulated entities are now solely responsible for the data privacy and security of investors, for the output of any AI tool they use (built in-house or bought), and for staying within the law. In plain words, “the AI wrote it” is not a defence. That makes Anthropic’s approve-before-send, log-everything design less of a nice-to-have and more of a requirement for anyone who tries to build the Indian version.

And someone will. The skills are published so firms can adapt them, and the connector pattern is open. I’d be surprised if an Indian wealth platform doesn’t ship something similar within the next year.

For Indian marketers in BFSI, insurance or pharma, the lesson is more immediate. The compliance-screening pattern (AI flags, a human decides, the log keeps the receipt) works for any regulated marketing team. You don’t need to wait for a finance plugin to set that up.

The verdict

For US advisers, Claude for Financial Advisors looks like a serious attempt at the right problem, with the right partners and a sensible approval model. Whether it saves the hours it promises is still unproven, and the first honest numbers will come from firms, not from the launch page.

For everyone else, it’s a preview. Connectors to the systems of record, skills built around the moments in a working day, a human approving anything that reaches a customer, and an audit log underneath. That’s the shape AI is going to take in marketing too.

When I wrote about Adobe’s Coworker last week, I ended on data, consent and approval workflows being clean enough for an agent to act on safely. This launch makes that point again from a completely different industry. If you’re in marketing ops, the homework is the same. List the five moments in your week that eat the most hours. Then check which of the systems behind them an AI could actually read today.

Quick answers

Related: Claude Now Has Its Own Browser. I Let It Work on My Site