- External Presence
- Reference Materials
A Guidebook to Claude for Financial Advisors
By The Impruve Team
Anthropic debuted Claude for Financial Advisors live on stage at Future Proof 2026, and Impruve was up there with them. Impruve CTO Danny Shannon ran the first live demo with Anthropic’s Drew Parker, then the two of them took questions from a room full of RIA leaders.
We told WealthManagement.com’s Davis Janowski, in his coverage of the announcement, that the out-of-the-box skills are only the start. What matters more for the wealth tech stack is MCP. Anthropic has set Claude up to connect to the systems advisors already use, which moves the industry toward an open garden. Drew hinted on stage that more MCP server announcements are coming.
This page collects the questions RIAs are asking about Claude for Financial Advisors and Claude Enterprise, starting with the ones from the debut. Most answers include the clip from the session, and every one has a short written version for anyone who would rather read than watch. We’ll add new questions as they come in.
Start here. When asked what implementation support RIAs can get for Claude Enterprise, Anthropic’s Drew Parker pointed the room to Impruve.
Watch all the clips on YouTube
The questions that came up
Click any question to jump to its short answer.
- What implementation support is available to RIAs for Claude Enterprise?
- Can firms put sensitive client data into Claude Enterprise?
- What new connectors and MCP partners are coming to Claude for Financial Advisors?
- Do I need zero data retention on my Claude Enterprise license?
- What does Claude Enterprise cost, and how do spend caps work?
- Can RIAs track spending in real time in Claude Enterprise?
- Which models should RIAs use for the workflows in Claude for Financial Advisors?
- Are connector tokens billed differently from chat tokens in Claude Enterprise?
- Can advisor groups keep their data separate in Claude Enterprise?
- Are connectors different between individual plans and Claude Enterprise?
- Can a firm customize and distribute skills to everyone in Claude Enterprise?
- How can teams share context and collaborate in Claude Enterprise?
- Why isn’t zero data retention required for sensitive data in Claude Enterprise?
- Can Claude Code use the same skills and connectors as Cowork?
- How do Claude Team and Claude Enterprise differ?
- How is learning from firm context different from model training in Claude Enterprise?
- How should RIAs handle AI hallucinations?
What implementation support is available to RIAs for Claude Enterprise?
The short answer. Anthropic’s go-to-market team is small relative to demand, so rather than trying to hire its way to scale, Anthropic trains partners with domain expertise to handle implementation, use case refinement, token optimization, and team onboarding. On stage, Drew pointed to Impruve as the partner for firms in the room.
Can firms put sensitive client data into Claude Enterprise?
The short answer. Yes. Anthropic doesn’t train on any data entered into a Claude Enterprise license, and each firm’s tenant is fully isolated. Prompts, the data Claude pulls through connectors, and Claude’s outputs all stay inside the firm’s four walls. Outside of Anthropic’s automated safety checks, nobody at Anthropic can open a firm’s chat history.
What new connectors and MCP partners are coming to Claude for Financial Advisors?
The short answer. On stage, Anthropic said it was aiming for a phase two launch in October with a long list of additional MCP partners. Firms that want specific tools included are invited to send feedback so those conversations can happen.
Do I need zero data retention on my Claude Enterprise license?
The short answer. No, and you wouldn’t want it. Without retention, Claude would have no memory, like a new hire you have to re-coach on every prompt. Because of data isolation, firms can retain data for as long as they like.
What does Claude Enterprise cost, and how do spend caps work?
The short answer. Unused seats incur no token charges, and tokens are billed at Anthropic’s public list rates, the same as the API. Spend caps are the control lever and can be set at the organization, group, and individual level. A practical start: set your monthly budget as the org cap, review spend on an ongoing basis, and raise it as new workflows prove their value. Admins can also review individual chat histories through the compliance API.
Can RIAs track spending in real time in Claude Enterprise?
The short answer. Yes. The Analytics API lets you pull spend programmatically in real time, broken down by who is spending what and on which surface of Claude: chat, Cowork, or Claude Code. There’s also a spend dashboard in the admin panel.
Which models should RIAs use for the workflows in Claude for Financial Advisors?
The short answer. The full menu of models is available, and the workflows run in a normal Cowork session. Start with Sonnet. It handles most rote automation workflows well, and the higher models are often overkill. Step up to Opus when a workflow needs more processing, accepting a bit more latency and cost. Knowing when to step up is a matter of human judgment.
Are connector tokens billed differently from chat tokens in Claude Enterprise?
The short answer. No. When Claude pulls data from a connector, those tokens are billed exactly the same way as tokens consumed in chat. All inputs and outputs are billed at the list rates published at claude.com/pricing. One cost tip from Danny: when you build custom workflows, tell Claude which connectors to pull from so it doesn’t waste tokens on tool calls to systems that aren’t relevant.
Can advisor groups keep their data separate in Claude Enterprise?
The short answer. Yes. Every seat, as well as the tenant overall, has data isolation. What one user enters into their Claude is private from other users. Connector permissions follow user credentials too, so if two advisors both connect Orion, each one’s Claude can only see the households that advisor can see in Orion.
Are connectors different between individual plans and Claude Enterprise?
The short answer. No, the same connectors are available on both. The caution is different: individual plans aren’t governed by the commercial terms of service, so connecting a firm system like Orion to a personal Claude plan is a bad place to be if the SEC comes asking.
Can a firm customize and distribute skills to everyone in Claude Enterprise?
The short answer. Yes. A firm can design its own plugin that bundles skills with firm-level context and customizations, keep it private to the firm, and push it out to everyone in the tenant with a simple install. Danny’s take: this is where a firm’s AI program pays off. Watch how people are creatively using Claude, spot the pockets of similar-but-different workflows, standardize them, and roll out the firm standard.
How can teams share context and collaborate in Claude Enterprise?
The short answer. With Claude Tag, you can add Claude to a Slack channel (and eventually Microsoft Teams) alongside the people who need to be there, and work with it directly in the channel. That’s how a workflow goes from one person in a Cowork session to a whole team.
Why isn’t zero data retention required for sensitive data in Claude Enterprise?
The short answer. Because of data isolation. It’s the same reason firms are comfortable entering PII into their CRM and other trusted systems: a walled-off tenant that’s specific to you, where nobody is training on the data, packaging it, or selling it. Danny’s view: there’s vendor risk with any vendor that holds PII, but Anthropic has taken the steps to make Impruve comfortable keeping that data under the terms Drew described.
Can Claude Code use the same skills and connectors as Cowork?
The short answer. Connectors, yes. The skills are built for Cowork, but they can be migrated; tell Claude you want to migrate the skills to work in Claude Code and it will walk you through the process. A general tip from both speakers: when you have a how-do-I question, ask Claude. It will usually figure it out.
How do Claude Team and Claude Enterprise differ?
The short answer. Neither plan trains on your data; both are governed by the commercial terms of service, so data isolation is standard on both. Claude Enterprise adds the admin and compliance layer: the compliance API, audit logs, custom RBAC controls, and other admin controls. Pricing works differently on each plan; see claude.com/pricing for current details. Danny added that compliance logging matters most once Claude starts writing to systems of record, so you have the record of what Claude changed and who approved it.
How is learning from firm context different from model training in Claude Enterprise?
The short answer. Training means Anthropic using your data to develop future model releases, which doesn’t happen. Context means teaching your own Claude how you want it to work, the way you’d onboard a new hire with your decision process for a given scenario. That context layer lives in your instance of Claude, isolated at both the seat and tenant level, and is never sent back to Anthropic. The flip side: a lesson one firm teaches its Claude doesn’t benefit anyone else, and even within a firm it’s isolated to the individual user unless you share it.
How should RIAs handle AI hallucinations?
The short answer. Hallucination rates keep falling as the models improve, but the principle doesn’t change: Claude creates the output and you own the deliverable. Review it the way you’d review work from a junior analyst before you put your name on it. Danny’s addition: generating content is cheap now, so the value is in the judgment. You’re the expert, your name is on the work, and it’s on you to steer Claude when it heads somewhere you wouldn’t.
Rolling out Claude Enterprise at your firm?
Impruve is the implementation partner Anthropic pointed to on stage. We work alongside RIAs and wealth management firms on use case refinement, token optimization, team onboarding and enablement, and the AI program that turns individual wins into a firm standard.
Recorded live at Future Proof 2026. Answers are lightly condensed from the on-stage remarks; the full session video is the source of record. Claude, Claude Enterprise, and Claude for Financial Advisors are products of Anthropic.