Claude 101 for Customer Success
A build-first guide to Claude for CS teams: the surfaces, the starter stack, and your first builds.
Most customer success teams already "use AI." In practice that means a CSM occasionally pastes something into a chatbot, gets a generic answer, and goes back to doing the work by hand. That is not fluency, and it is not the ceiling. This is the foundation that closes the gap: what Claude actually is, the handful of surfaces that matter for CS, and the starter stack you can set up this week so that by the end of it you have built something, not just read about it. It is written for both the CS leader standing this up for a team and the CSM standing it up for themselves.
- "Using AI" by pasting into a chatbot is the starting line, not the finish. Real value comes from connecting Claude to your accounts and building repeatable workflows.
- Claude is an assistant plus a set of surfaces. The four that matter for CS: Projects (persistent context), connectors (your live data), Skills (reusable builds), and the work surfaces (Cowork, Claude in Excel).
- Your foundation is the CS Starter Stack: one Project, a context file, one connector, and one Skill. Set up this week, it changes every answer Claude gives you from generic to yours.
- Settle the data question first, then build. Once the stack is in place, Claude 201: The CS Build Kit (coming next in this series) hands you a copy-paste context file, ready-to-run prompts, and a real Skill to build on top of it.
Claude is an AI assistant from Anthropic that customer success teams use to draft, summarize, analyze, and automate account work. Used well, it is not a chatbot you occasionally visit but a workspace connected to your accounts that produces the first version of nearly everything a CSM writes and reads.
What Claude actually is, and what it is not
The most common mistake in CS is treating Claude as a smarter search box: ask a question, copy the answer, leave. That gets you generic output because the model knows nothing about your accounts, your playbook, or your voice. It is the equivalent of asking a brilliant new hire for help on their first morning, before they have read a single account.
Claude is two things working together. It is a capable assistant that can reason, write, and analyze. And it is a set of surfaces that let you give that assistant your context and your data, and let you package the useful patterns so they run the same way every time. Fluency is not about knowing clever prompts. It is about setting up those surfaces so the assistant works inside your world instead of in the abstract.
That distinction is the whole of this guide. Everything below is about moving from "I asked a chatbot" to "I built a workspace that knows my book of business."
The CS Starter Stack
The four things every CS team sets up first. Each one moves Claude further from generic and closer to a teammate who knows your accounts. Stand all four up and you have a foundation you can build on for months.
A Project is a persistent workspace that remembers the context you load into it, so you stop re-explaining your accounts in every chat. Everything else in the stack lives here. Create one and call it something like "My Book of Business."
The document, loaded into the Project, that tells Claude who you are and how you work: your playbook, your ICP, the product's use cases, your QBR template, and writing samples so it picks up your voice. This is what turns generic output into output that sounds like you and lands in your world.
A connector wires Claude to where your data actually lives: your calendar, email, CRM, or customer success platform. With one live, Claude answers across your real book instead of a snippet you pasted, and can prep your day or summarize an account on request.
A Skill is a reusable, named build that tells Claude how to run a specific task the same way every time. Your first one can be simple, a call-prep brief or an account summary. It is the seed of every workflow you will build later.
You do not need all four perfect on day one. You need all four in place. The Project holds the context, the context file makes answers yours, the connector brings in live data, and the first Skill proves you can package a pattern instead of re-prompting from memory. That is the foundation the rest of the series builds on.
The entire starter foundation is four pieces a CSM can set up in an afternoon: a Project, a context file, one connector, and one Skill. Most teams overestimate what AI adoption requires and never set up the basics that make every later step work.
Surface by surface, what each is for
A short tour of the surfaces a CS team actually touches. You will not need all of them at once. Know what each is for so you reach for the right one.
Projects
A Project is where fluency compounds. Instead of starting cold every time, you work inside a space that already holds your playbook and your accounts. Use one Project per book of business, or one per major account if you run a small number of large customers. The rule of thumb: if you find yourself pasting the same background into chats, that background belongs in a Project.
Connectors
Connectors are how Claude stops guessing and starts answering from your real data. Connect your calendar and email and it can prep you for the day. Connect your CRM or your CS platform and it can answer "which accounts are quiet and renewing soon" without you clicking through five tabs. Start with one connector, learn what it exposes, and add others as you trust the pattern.
Skills
A Skill turns a good prompt into a dependable tool. Once you have figured out how to get a strong renewal narrative or a clean QBR outline, a Skill captures that so anyone can run it the same way, including you next month when you have forgotten the exact wording. Skills are also how a workflow becomes a team asset: a documented Skill can be shared and run by a peer.
Cowork and Claude in Excel
These are the work surfaces for people who are not engineers. Cowork lets you automate the file-and-document side of your work, turning a messy set of inputs into a finished deliverable. Claude in Excel is where your usage data, health scores, and renewal lists already live, so it is often the fastest place to turn raw numbers into a read. For the technically inclined, Claude Code goes further, but most CS work never needs it.
Choosing a model
Claude comes in tiers. Use the balanced everyday model for daily drafting and account work, reach for the most capable model on genuinely hard reasoning like untangling a confusing at-risk account, and use the fast model for high-volume, low-stakes passes. Most CS work sits comfortably on the everyday model, so do not overthink this.
Your first five prompts
Fluency is built on reps, so start today. Run these against real accounts, not hypotheticals, ideally inside the Project you just created.
The account read. "Here is the latest activity, support history, and usage for this account. Tell me how healthy it is, what is most likely to go wrong before renewal, and the one thing I should do this week."
The next-move drafter. "Here is the last QBR summary and the most recent thread with this customer. Draft three options for my next outreach, ranked by what I should do first, and tell me what I am missing."
The meeting prep brief. "I have a call with this account in an hour. Pull together a one-page brief: where we are, open items, risks, and three questions I should ask."
The renewal story. "Based on this account's usage trend and stakeholder notes, write the renewal narrative I would tell the customer, and separately the honest version I would tell my own forecast."
The time audit. "Here is how I spend a typical week as a CSM. List the recurring, rules-based tasks and rank them by how many hours I would get back if AI drafted them first." That last one tells you what to build first.
The data question comes first
Before any of this touches a real customer, settle the data question. Confirm with your security or legal team what customer data is allowed into your AI tool. Use a business or enterprise plan with the right data-handling terms rather than a personal account. Default to de-identified examples when you are unsure, and learn what your connectors expose before you turn them on.
This is not a footnote. It is step zero. A team that builds real fluency on top of a quiet policy violation has not saved time, it has created a problem that surfaces at the worst possible moment. Done right, the data question is answered once, written down, and never slows you down again.
Fluency is not knowing clever prompts. It is setting up the surfaces so the assistant works inside your world.
Where this goes next
The starter stack is the foundation, not the finish. Once a Project, a context file, a connector, and a first Skill are in place, you are ready to build something real and repeatable. Claude 201: The CS Build Kit, coming next in this series, takes this foundation and hands you a copy-paste context file, four full prompts, and a real Skill, so the stack stops being something you set up and starts building work for you. It is one rung on the larger path mapped in The CS Leader's Guide to Mastering Claude, the best place to see where this leads and the program it builds toward. And if you are still wondering who on your team should lead this, the answer is usually already on it.
Your move
- Settle the data question, then create the Project today. Confirm what customer data is allowed in, then create one Project and load a context file: your playbook, your voice samples, and three real account briefs. This is the single highest-leverage hour you will spend.
- Run the five prompts on real accounts this week. Reps on real work, not demos, are how the habit takes. Pay attention to which outputs you actually keep.
- Build one tiny Skill. Turn your best prompt into a reusable Skill, even a simple account summary. Proving you can package a pattern is the doorway to everything that comes next in this series.
FAQ
Ready to build your CS foundation on Claude?
See how ClientSuccess teams put these principles to work.