How to Use ChatGPT for Customer Success: A Guide for CSMs

We’re all being told to “use AI more,” but what does that really mean?

Should you only trust ChatGPT with your most basic questions, or let it run point for your biggest accounts? It feels like everything changed when ChatGPT first took the world by storm in 2023, but even now it isn’t a one-size-fits-all solution. At least not for every CS task.

We’ve asked the CS community and our customers how they’re using AI effectively. We’ll share our findings below to show you how CSMs can use ChatGPT to cut hours from their workflow and deliver a better customer experience at scale.

What is an LLM?

AI may seem like magic, but it isn’t; it’s just machine learning. 

ChatGPT runs on a large language model (LLM); this is the real breakthrough that brought AI into the mainstream, so it helps to understand how LLMs work before continuing. 

LLMs are machine learning models trained on massive collections of text and images. And training these models takes an unfathomable amount of text, which is why OpenAI has a content partnership with Reddit and other platforms.

LLMs work by making a statistical guess at the most likely response for a given prompt, even down to which word is most statistically likely – and that is the reason ChatGPT “hallucinates.” ChatGPT doesn’t “read” your prompt and “think” the way you do. It takes in your prompt, breaks down the words, and makes an educated guess at the most likely response. 

Why go into this level of detail? This is why generic prompts like “write me an email” without context produce bad, bland, or incorrect responses. AI needs guardrails. Below, we’ll show you exactly how to build those guardrails in ChatGPT and how better context can lead to better outputs.

4 Ways to Use ChatGPT Beyond Simple Prompting

Getting the most out of ChatGPT starts by going beyond basic prompting. The chat version is great, but it doesn’t match OpenAI’s full capabilities.

Better context will help, but you should also take advantage of ChatGPT’s more powerful agentic features:

• Custom GPTs: Writing the same prompt over and over again? Automate it with a custom GPT. GPTs are text files with specific instructions and context that can guide ChatGPT on how to think and respond.
• ChatGPT Work: OpenAI’s counterpart to Claude Cowork; an agentic productivity tool designed for non-technical people to queue up complex, multi-step tasks for AI. ChatGPT Work can access and create files on your desktop, and even interact with your browser. 
• Codex: For those who are a bit more tech-savvy, OpenAI offers an agentic coding platform similar to Claude Code called Codex. Like Claude Code, though, you don’t need to be a developer to use it. You can write plaintext instructions, and ChatGPT handles the rest.
• MCP (Model Context Protocol): An MCP lets ChatGPT connect to your favorite tools and access their data. You can prompt ChatGPT like you normally would, and trust that it has secure access to real-world data and can perform complex tasks on top of it.

Together, these four tools let you access real-world data, customize ChatGPT's responses, and ultimately get better outputs and save time. Asking ChatGPT to generate a follow-up email is fine, but these advanced methods are the difference between using AI to save five minutes from your day and using AI to shave off five hours each week. 

5 Ways CSMs Can Use ChatGPT Without Sounding Like a Robot

Let’s get practical. We’ve broken down five ways that CSMs can use ChatGPT with guidance on how to not sound like an AI chatbot:

• QBR prep
• Call follow-up
• Churn management
• Spot expansion opportunities
• Renewal management

How to Use ChatGPT for QBR Prep

Sure, you need to show up and be on your A-game for a QBR, but that doesn’t mean you need to shoulder the load of meeting prep all on your own.

Pulling data, catching up on an account, creating slides from a template – these are all great tasks for AI to automate. Set up a custom GPT that takes on the following:

• Renewal date
• Last interaction
• Open action items
• Usage highlights
• Red flags
• Necessary context

And have ChatGPT Work create a slide deck from the template. The key to not sounding robotic is to start with a solid template and use ChatGPT to pull data and insights. ChatGPT excels at data analysis, so this should be in its wheelhouse.

Check out our guide to using AI for QBR prep for a few more tips and prompts.

How to Use ChatGPT for Call Follow-Up

Ever ended a call with the best of intentions only to find that your follow-up email isn’t written until four hours after the fact? 

Don’t worry, it happens. ChatGPT can help here too, though. AI can’t replace you, but it can be a great assistant here by taking in your call notes, a transcript, and any relevant account notes and queuing up follow-up emails for you without lifting a finger.

If you use an AI-powered Customer Success Platform like Vitally, you’ll get access to an AI notetaker that reviews your calls along with customer and subscription data to generate tasks and next steps for every account on the fly. To avoid those obvious AI-isms in AI-generated text, give ChatGPT plenty of context about your actual voice. It helps to include examples of past emails you’ve sent, or even a list of phrases you like to use or not use. 

How to Use ChatGPT to Identify Churn Risks

Most CSMs have a sixth sense when an account is about to become a churn risk, but when your book of business crosses 100, good luck keeping an accurate read on that from week to week.

AI can help again by crunching numbers for you at scale. Grab your account and subscription data, renewal dates, support sentiment data, health scores, and product usage data, and feed that to ChatGPT Work or Codex for an advanced AI-powered churn analysis. 

If you have a Customer Success Platform with an MCP like Vitally, you’ll have all of your customer, revenue, support, and product data in one place and can then run automated health checks each week with ChatGPT. 

How much time could this save you each week? If you ask Larry Hogan, VP of Customer Success at Triptease, up to 30% of a CSM’s time:

“Before Vitally, if I asked a CSM, ‘What’s your churn rate this quarter?’, the answer was, ‘I don’t know, I’ll tell you at the end of the quarter.’ If your bonus is based on a number and you don’t know that number until the end of the quarter, how can someone maintain their sanity?… The concept of being a truly strategic CSM was impossible before. With the AI features in Vitally, they now actually have the time and visibility to do the work they were hired to do.”

How to Use ChatGPT to Spot Expansion Opportunities

If you're providing your customer, revenue, and product data to ChatGPT, then it can help you spot expansion opportunities as well.

You can have ChatGPT Work or Codex run a weekly task to scan accounts for those close to hitting usage limits, expanding headcount, or with the most favorable support interactions, then craft perfectly personalized messages for each. 

Just as with the follow-up emails, give ChatGPT samples of actual emails you’ve sent so that it can generate outputs that match your tone and writing style. It won’t be perfect, but even 90% gives you a massive head start. 

How to Use ChatGPT for Renewal Management

More than ever, CSMs must steer renewals across an expanding book of business. Managing that manually is a fast way to lose your sanity, but fortunately ChatGPT can help immensely. 

By giving ChatGPT access to your health score and NPS data (or using a Customer Success Platform with an MCP), you can queue up automated tasks for the following:

• 120 days out: Raise early churn signals and flag at-risk accounts 120 days out from renewal. You can and should customize the signals that trigger these alerts.
• 90 days out: Have ChatGPT review call transcripts for any signs of buying friction. This could be budget pushback, hesitation about expansion, or requests for ROI.
• 30 days out: Run a scheduled task for any final risks that point to churn. This could be things like lapses in communication, contract revisions, or delayed approvals

Instead of pulling data and doing this manually, you can automate almost everything, so you show up prepared with a personalized game plan for every account.

ChatGPT vs Claude: Which is Better?

ChatGPT vs Claude is the latest “Apple vs PC” debate. Does it really matter?

Honestly? No. At least, not as much as OpenAI or Anthropic would have you believe. 

ChatGPT was first to market in 2023, but since then, Claude’s models have gotten better, and Anthropic has added new tools like Claude Code, Claude Cowork, and Claude Design. 

Anecdotally, you’ll find claims that Claude is better for deep work and writing, while ChatGPT does better for data analysis. Well, if you paid close attention to the rest of the article above, you’d notice that there are plenty of tasks that are data-intensive and plenty that involve a fair bit of writing. Try them out and see which fits your needs better. 

Remember, better outputs come from better context and data, not necessarily from which AI provider has the most up-to-date model. That’s where a Customer Success Platform can make either ChatGPT or Claude truly effective. By matching AI with the full context of your customer data, you’ll generate outputs that actually save you time rather than adding another thing that needs babysitting to your plate.

Download 23 AI Prompts for Customer Success

ChatGPT does so much more than answer basic questions. With Work and Codex, you can scale entire workflows on top of your data, freeing you up to do what you do best: be human for your customers.

If you’re unsure of where to start, don’t worry. We compiled a list of our favorite prompts to automate your CS workflows, from spotting churn risks to following up after meetings. Save your favorite prompts, create a custom GPT, and see how much time you can save across all of your accounts.

Click the button below to get your free copy.