How to Set Up Claude for Your Advisory Practice (A Step-by-Step Guide)

Claude has two main modes. Understanding when to use each one saves you time and frustration.

Chat is for thinking. You're having a conversation. Brainstorming ideas, asking questions, working through a problem, getting feedback on something you've written. Chat is back and forth. You prompt, Claude responds, you refine.

Use Chat when you want to: bounce ideas around, ask "how should I approach this," get feedback on a draft you paste in, research a topic, or think through a client scenario.

Cowork is for doing. You describe the outcome you want and Claude executes it. It can read your files, create documents, browse the web, organize folders, and run multi-step workflows without you prompting at every step.

Use Cowork when you want to: create a batch of content from your notes, process meeting notes into follow-up emails, generate a quarter's worth of LinkedIn posts, build documents, or run any workflow where the output is a finished product.

The simple rule: if you're thinking, use Chat. If you need something built, use Cowork.

Most advisors default to Chat for everything because it feels familiar. But once you start using Cowork for execution tasks, you'll wonder why you ever wrote prompts for things that should just get done.

Claude has several plan tiers. Here's what matters for advisors.

Free plan. You get access to Chat with Claude Sonnet. Good for trying things out. No Cowork access. Limited usage.

Pro plan ($20/month). This is where most advisors should start. You get Chat and Cowork, access to all models (Sonnet and Opus), Extended Thinking, and the ability to install Skills and plugins. There are usage limits. If you're running Opus with Extended Thinking on every task, you'll hit those limits mid-week. More on that below.

Max plan ($100 or $200/month). Significantly higher usage limits. Worth it if you're using Cowork heavily every day or running complex workflows that eat through tokens. Most advisors won't need this right away, but if you find yourself hitting Pro limits regularly, it's the upgrade path.

Start with Pro. See how your usage feels for a couple weeks. Upgrade if you need to.

This is where most generic advice gets it wrong. The common recommendation is "always use Opus 4.6 with Extended Thinking." That sounds good but it burns through your usage allocation fast and isn't necessary for most tasks.

Claude Sonnet 4.6 is your daily workhorse. Fast, capable, and efficient with your usage limits. Use it for writing LinkedIn posts, drafting client emails, brainstorming content ideas, editing drafts, creating social media content, and most day-to-day communication tasks. This handles 80% of what advisors need.

Claude Opus 4.6 is for complex work. Use it when you need deep analysis across multiple documents, complex financial scenario reasoning, building comprehensive frameworks or systems, long-form content that requires maintaining consistency across thousands of words, or any task where you notice Sonnet isn't quite getting the nuance right.

Extended Thinking adds extra reasoning time before Claude responds. Valuable for complex multi-step analysis, strategic planning, or when you're asking Claude to consider many variables at once. Not needed for drafting a LinkedIn post or writing an email. Turn it on when the task is genuinely complex. Leave it off for routine work.

The practical approach: start every session on Sonnet. If the output isn't meeting your expectations for a specific task, switch to Opus. Turn on Extended Thinking only when you need deep reasoning. This keeps your usage sustainable across the month instead of running out by Wednesday.

Before you open Cowork for the first time, create a folder on your computer. This is where Claude will read your context and save its output.

Here's a structure that works well for advisors:

Claude-Work (your master folder)

Inside that, four subfolders:

About Me. This is where Claude learns who you are. Your voice guide, your practice context, your preferences. More on what goes in here below.

Projects. One subfolder per active project. If you're working on a quarterly newsletter plan, that's a project. If you're building a content series for a specific niche topic, that's a project. Each project gets its own space so Claude can focus on the relevant context.

Templates. Your best work as reusable structures. A great LinkedIn post format you want to replicate. Your preferred email structure. A client letter template you like. Claude references these when creating new content so it follows your proven patterns.

Outputs. Where Claude saves finished work. This keeps your deliverables separate from your context files so you can easily find and use what Claude creates.

When you open Cowork and point it at this folder, Claude can read everything inside it. That's how it knows who you are without you explaining it every time.

Inside the About Me folder, create three files. These are plain text files saved with a .md extension. You can create them in any text editor (Notes, TextEdit, Notepad, whatever you use).

about-me.md

This is not your resume. It's what you do day to day and what Claude needs to know to help you effectively.

Include: your role (financial advisor, CFP, whatever your title is), your niche (who you serve specifically), what you spend most of your time on (content creation, client meetings, planning, prospecting), the platforms you use (LinkedIn, Substack, email), and your firm's general compliance approach.

Keep it conversational. Write it like you're explaining your practice to a smart friend who's going to help you with projects.

my-voice.md

This is your voice guide. If you've already built one (I teach how in my newsletter), paste it here. If you haven't, at minimum include: your tone (direct, warm, analytical, encouraging), phrases you like to use, phrases you never want to see, 2 to 3 samples of your best writing, and how you naturally handle the education vs. advice line.

This file is the most important one. It's the difference between Claude producing generic output and producing content that sounds like you.

my-rules.md

This is how you want Claude to work with you. Things like:

Ask clarifying questions before starting complex tasks. Show me a plan before executing multi-step workflows. Never delete files without my explicit approval. Always include compliance guardrails in content drafts. Use Sonnet unless I specifically request Opus.

These rules apply to every session automatically once you set them up.

This is a one-time setup that applies to every Cowork session going forward.

Go to Settings, then Cowork, then Edit Global Instructions. Paste something like:

"I'm [Your Name], a [Your Credentials] who works with [Your Niche]. Read my context files in the About Me folder before every task. Ask clarifying questions before executing complex requests. Show a plan before acting on multi-step workflows. Never delete files without my explicit approval. All client-facing content must include compliance guardrails: educational framing, no guarantees, no specific product recommendations, qualifying language where appropriate."

Customize it for your practice. The key elements are: who you are, read my files first, ask before acting, and maintain compliance standards.

Once this is set, your prompts can be short because Claude already has the context. Instead of writing a paragraph of instructions every time, you can say "draft a LinkedIn post about inherited IRA rules for my niche" and Claude knows your voice, your audience, your compliance patterns, and your preferred format.

This is important and most AI guides skip it entirely.

As of early 2026, neither the SEC nor FINRA has issued regulations specifically addressing the use of AI for content creation by advisors. However, FINRA's existing guidance (including Regulatory Notice 24-09) makes clear that existing rules, which are designed to be technology-neutral, continue to apply when firms use AI tools.

What does that mean practically?

Your recordkeeping obligations don't change because you used AI. If you draft a client email with Claude's help, the final email you send still needs to be archived the same way you'd archive any client communication. If you publish a LinkedIn post that Claude helped you write, it still needs to go through whatever review and archiving process your firm requires for social media content.

Claude's conversation history is stored locally on your device (in Cowork) or on Anthropic's servers (in Chat). But that is not the same as compliant archiving for your firm. Do not rely on Claude's conversation history as your regulatory archive.

Here's what I'd recommend:

Treat AI-generated content the same as any other draft. You review it, you approve it, you publish it, you archive the published version through your firm's normal process.

Document your AI workflow. Write a brief description of how you use AI in your practice and keep it in your compliance file. Something like: "Content is drafted with AI assistance using my voice guide and compliance guardrails. I review all output for accuracy and compliance before publishing. All published content is archived through [your archiving system]."

Talk to your compliance team. Don't hide your AI use. Be proactive about explaining your process. Most compliance professionals are receptive when you show them a thoughtful, documented approach.

It's also worth noting that regulatory guidance in this area is evolving. FINRA's 2026 Regulatory Oversight Report added a new section on generative AI, and they're clearly paying attention to how firms are using these tools. State regulators and the SEC may move at different paces on this. Stay informed, and when new guidance comes out, update your workflow documentation accordingly.

The regulatory framework will catch up to the technology eventually. The advisors who have a documented, thoughtful process in place now will be well positioned when it does.

Once everything is set up, here's how to run your first session:

Open the Claude desktop app. Click the Cowork tab. Point it at your Claude-Work folder. Type something simple:

"Read my About Me files and confirm you understand my practice, my voice, and my preferences."

Claude will read through your context files and summarize what it knows. Review the summary. If anything is off, correct it. If it looks good, you're ready to work.

Then try a real task. Something like:

"Draft two LinkedIn posts for this week about [topic]. Use my voice guide. Include compliance guardrails. Save them to my Outputs folder."

That's it. No long prompts. No re-explaining who you are. Claude already knows. Your setup did the work so your prompts don't have to.

If you've made it this far, you have everything you need to set up Claude for your advisory practice. Block 30 minutes. Create the folder. Write the three context files. Set your global instructions. Run your first session.

The advisors who get the most out of AI aren't the ones with the fanciest prompts. They're the ones who took 30 minutes to set things up properly so every session starts from context, not from zero.

You're already ahead by thinking about this. Now go build it.

Amplify for Advisors teaches financial advisors how to use AI with a voice-first, compliance-aware approach. Every issue includes frameworks, prompts, and systems built for how advisors actually work.

Subscribe for free at amplifyforadvisors.substack.com or explore the full system at amplifyforadvisors.ai.