How to Build a “Second Brain” for Customer Success

As good as LLMs are, we all know they are far from perfect. Just check out some hilarious gaffes shared on LinkedIn after each new model release if you need any more convincing.

The fact is, there’s a lot of context and real-world information tucked away in your brain that AI lacks. You could spend an hour writing the perfect prompt, but that’s going to get old quickly if you are an AI power user. 

To speed things up and improve output quality, you need a “second brain” to guide AI. Linnea Olson and Nina Wilkinson from ScaleUp CS joined us for a recent webinar on this very topic. In this guide, we’re distilling what Linnea and Nina shared to explain what a second brain is, how to create one, and how using one will improve the outputs you get from AI tools. 

What is a Second Brain?

A second brain (a concept popularized by Tiago Forte) is a system that lives outside your head, capturing, organizing, learning, and surfacing everything you know, bringing it back to you when you need it.

This was originally a productivity tool for building a personal knowledge system, but it has been adapted for AI to provide necessary context, guardrails, and instructions.

For most CS teams, their context lives all over the place. Customer data is in their CRM, health scores are calculated in spreadsheets, and product data is sent from a BI tool. When all of your data is disjointed like this, you’re going to have an incredibly hard time combining and normalizing it to get something useful from it. 

If you bring all that information into your AI stack, the system will learn who you are and will be able to help you build all of your markdown files and your organized files into a system that AI can actually reason over and use to build out content that you need.

How to Set up Your Second Brain

Every CSM who’s opened Claude or ChatGPT and expected a magic fairy godmother-level of perfection quickly realized that AI is far from infallible. 

That’s because there’s a fundamental difference between how you think and how AI “thinks.” When someone asks you a question, your brain draws on years of experience and firsthand knowledge while reacting to brand-new inputs in real-time. LLMs (large language models) are trained on text and image data and perform a series of mathematical calculations to predict the most statistically likely response.

These tools are getting better, but they will still make mistakes; you can improve your chances of getting a valuable response with the following prep work:

• Markdown files
• PARA structure
• Tone/voice materials

Markdown Files

A markdown file (also called an MD file) is a plain-text format used to provide instructions for an LLM. 

Why does it matter for AI? Claude, in particular, reads and writes markdown files natively, so these instructions are easy for it to follow. Markdown files also survive copy-and-paste across Slack, Notion, Google Docs, and your CSP. And markdown files are lightweight, future-proof, and version-control-friendly, making them the perfect choice for an ongoing record of instructions.

Claude already runs its own “system prompt” before every conversation, so creating your own markdown file is a way to guide Claude’s responses to what you are specifically looking for. Check out our guide to creating Claude Skills for Customer Success for a full walkthrough on creating custom markdown files (and a downloadable set of markdown files to use).

Identify the core jobs-to-be-done in your role – the routine tasks you want to automate – and write instructions for how to complete them, then move on to the next step.

PARA Structure

If you’re a productivity nerd, you’ve probably heard of Tiago Forte’s PARA structure. We find that it helps a lot when organizing context and files for AI.

PARA is an acronym for a basic organizational structure that puts every piece of information you have into one of four buckets, allowing your AI agents to navigate that information autonomously without you having to consistently feed it. Those buckets are:

• P - Projects: The things you are actively working on, like a QBR deck.
• A - Areas: Your ongoing responsibilities, like renewal management or churn mitigation.
• R - Resources: The materials that you've built out in the past, such as frameworks, templates, and playbooks.
• A - Archives: Previously created content for AI to piggyback off of. This could be completed QBR decks, onboarding SOPs, whatever you need to train the LLM. 

We recommend putting everything into your PARA structure that seems relevant. The more source material Claude/ChatGPT has, the better its outputs will be. Feel free to tap other departments, like marketing and sales, for battle cards and case studies.

Tone/Voice Materials

The last step is to match the above context with materials that reflect your voice/tone to guide the AI as it generates content.

To do that, you just need to feed your QBR decks, onboarding training calls, marketing materials, one-pagers, and cross-functional assets, such as case studies and white papers, from your marketing team. You want the AI to learn your tone, voice, and perspective so it can produce work that sounds like you. 

This will give you a good base to work from for customizing Claude/ChatGPT to create the exact output you’re looking for and to sound just like you (or at least close to you). If you want to mix and match models, Nina mentioned using Obsidian to store notes in Markdown format and pitting Claude and Gemini against each other to improve the outputs (and to watch the models get a bit sassy with each other).

This is a lot of work, but trust us that it is well worth the effort.

It also helps to have a Customer Success Platform that brings all your data into a usable, standardized format. For example, Vitally brings in customer, support, product, and revenue data under one roof, allowing you to collect information into shareable dashboards, forecast revenue, spot churn risks, and build fully customized health scores. 

With your second brain and Claude, or an AI-powered Customer Success Platform like Vitally, you have everything you need to capture your customer information and act on it.

How to Use Your Second Brain for Customer Success

Without a second brain, you’re looking at extra back-and-forth with Claude to tweak and edit every response it generates. That’s valuable time lost to managing a tool that is supposed to take work off your plate, not add to it.

Savvy CSMs take the time to work through the above prep work because they know it will lead to more specific outputs tailored to them, rather than the generic outputs AI tools are designed to provide by default. 

A good second brain file allows you to do things like:

• Spot patterns in your data by recovering buried context and information across tools.
• Match your tone and writing style for every output.
• Recognize patterns over time, which can help spot churn risks or upsell opportunities
• Collect data and stakeholder context into fully designed QBR decks

In other words, it can automate some of your most time-consuming tasks to free you up for the important relationship-building work that AI could never replace.

Need Help With Those Markdown Files? Download These Claude Skills

Gathering source materials and integrating your apps is the easy part; you still need those markdown files to instruct AI on what to do and how to respond. 

Fortunately, we wrote them for you. Click the button below to access our Claude Skills library, where you can browse available markdown files written specifically for different CS tasks. Automate QBR prep, call follow-up, churn management, and much more. 

Click the button below to get a link to the files, upload them to Claude (either in the browser or in Cowork), and start automating your work so you can focus more on building relationships with your top accounts.