How OpenAI's customer success team uses AI (strategies)

I was the first customer success hire at OpenAI, and stepping into that role back in September 2023 gave me a front-row seat to something interesting. 

Everyone talks about AI as this productivity machine, this tool that automates the boring stuff and speeds up your day. And it does all of that.

But the thing I keep coming back to, the thing I want every CS professional to sit with, is how AI actually makes us more human in the way we work with customers. It frees you up to build stronger relationships, understand your customers more deeply, and show up to every conversation genuinely prepared to help.

Let me walk you through how I think about all of this, because I know how overwhelming the AI conversation can feel right now.

How customer success teams can use AI

Before we get into workflows and strategies, it helps to organize your thinking.

There's so much noise around AI that a lot of people feel lost before they even start. So I broke it into three distinct buckets, and once you can see them clearly, the whole thing becomes much easier to work with.

Keeping these three buckets separate genuinely helps – and at OpenAI, we use all three.

10 practical ways to use AI in customer success

In this guide, we break down 10 practical, real-world ways CSMs can use AI right now to deepen customer understanding, scale proactive support, boost productivity, and level up their strategic value.

Customer Success CollectiveGrace Gupta

1. Using AI to prepare for customer calls

Let's start with that first bucket, because this is where I want you to anchor your thinking.

At OpenAI, we use ChatGPT Enterprise constantly. It's a familiar interface for anyone who's touched ChatGPT before, and it drives real efficiency for us.

We use it to prep for customer calls. We do quick, efficient account research. We understand the history of an account, what the customer's business does, what their objectives are, and how AI can ladder into helping them accomplish their goals, all before I even join the call. When you show up to that conversation already carrying that context, the whole engagement improves.

The reason preparation is the real differentiator

This is a huge block of work that every CSM and every CS leader should be leaning into. Being better prepared for calls. Being more strategic on those calls. You can now use AI to walk in far more ready than you ever could before, and that alone will level up the quality of the conversations you're having.

2. The AI already sitting inside your tools

The second bucket is the AI showing up inside the tools you're already using.

Slack AI can summarize channels for you. And at OpenAI, we use Gong, which records our calls and produces transcripts afterward. That's useful on its own, but now Gong can summarize the call, pull out action items, and generate an email back to the customer with next steps at the click of a button. That can easily save you 30 minutes of summarizing notes and writing follow-ups.

There's a funny quirk I've noticed with that email feature, and I'll admit I'm a little obsessed with it.

Sometimes the action items will list the name of your conference room. So it'll say something like "conference room number seven has this action item," and you have to go in and correct it because that was actually you talking. It's a small thing, but it's a great reminder to really understand the tools your company is already investing in and figure out how to use them for your own efficiency.

3. Becoming your own analyst

Now let's talk about one of the genuine superpowers of generative AI, which is making more data-driven decisions.

I've been in CS leadership roles where we were sitting on mountains of customer data, and I'd have questions I desperately wanted answered. Where's our growth actually happening? Is it in a particular segment, a particular country, or a particular type of customer with certain characteristics? 

I'd have hypotheses, but because the data lived across different systems in huge datasets, I couldn't just answer them myself. I'd open a BizOps ticket, ask for help, wait, go back and forth, and eventually get an answer. I don't have to do that anymore. 

Connecting your data sources

I can pull data down from Salesforce, pull adoption and usage data from another system, and sometimes pull from a third system too. Then I upload all of it into ChatGPT Enterprise and run the kind of complex analysis I never could have done on my own. 

Total candor here – some of it I couldn't have done even with a dedicated BizOps person. The ability of generative AI to handle these complex mathematical and analytical tasks is genuinely powerful.

What I love most is that it often surfaces answers to questions I wouldn't have thought to ask. We can be our own analysts now, asking tougher questions – and I can do it myself in a couple of minutes, as long as I know where the data lives.

We've started connecting our ChatGPT Enterprise instance directly into tools like Salesforce, so I don't even have to pull spreadsheets anymore. I've built what we call a GPT, which holds my prompt and the analytical questions I want to run, and it connects to my data sources through an API on the back end. So the data's just there, ready.

That's a huge unlock for productivity and for being genuinely data-driven.

Always lean into the data

If you're building your career as a CSM, this is some of the best advice I can offer. Learn how to create a narrative with data, how to influence people with it, and how to make the right decisions for the business and for your customers using it. The best CSMs I've worked with all had this ability. 

So take the AI tools you can access, think about the questions that have been sitting in the back of your head that you never had time to explore, and run them. Then walk into your next meeting and impress your manager with thoughtful analysis, or convince a customer to unlock a new department or region by showing them the impact and the trends in their own data.

Letting AI challenge your thinking

I picked up a technique from one of our product managers at OpenAI, and it's changed how I use these tools. He's a superuser, as most of us here are, and he shared something simple but powerful.

When you're doing account research or drafting a strategy for an internal QBR, don't just take the response and move on.

Ask the AI what three questions you didn't think to ask. Ask it to give you three ways it would push back against your proposal. Using AI to challenge you is genuinely valuable.

I now regularly ask it to push me further, to tell me what I've missed, or to argue against my own plan so I can anticipate where people might hesitate.

The job security question

This connects to something I often hear from CSMs in my network: People can be afraid of AI. 

Is this thing going to take my job? I understand the worry. But the more useful question is which tasks AI can take off your plate, and how you then position yourself to take on more valuable work. 

Maybe you're no longer writing 15-minute follow-up emails because you can do that with one click. Great. Now you have an extra hour across your week for deep, data-driven analysis. That analysis unlocks revenue in your customer base, or it helps your product team build a better feature because you can show them the data. That's how I encourage everyone to think about it.

And here's something that surprised me. I feel like AI is actually making me smarter, because it pushes back on me. When it answers a prompt, it often shows its reasoning, and if it doesn't, you can ask it to walk you through its thought process.

Used the right way, I find myself having sharper ideas and anticipating objections more naturally, even when I'm not sitting in front of the tool. So treat AI as a way to upskill yourself, with or without it in front of you.

Personalizing communication at scale

Let's talk about tailoring your communication to individual customers, because this is where preparation and personalization meet. There are a couple of big use cases here. 

The individual customer conversation

Consider the one-to-one work you do as a CSM, whether that's a call, outreach, or a follow-up. You want to know:

• How well are you leaning into what actually matters for that specific customer? 
• Where are they in their journey? 
• What does their usage look like? 
• What industry and region are they in? 

The more of that context you feed in, the better the result.

Internally at OpenAI, we've built GPTs for exactly this. I can open one up five minutes before a call, type in the customer's name, and it spits out their business objectives, what they said on their latest earnings call or in recent articles, and how our products can help them hit those goals. 

When you then get on the call with a C-suite leader, you can speak directly to what their CEO and board care about. That's how the best CSMs earn the right to real partnership.

Communication at scale

Our digital success team sends emails to customers, and our support team handles tickets. AI lets us take customer behavior, usage, preferences, and interaction history into account when crafting the next message, at a level of personalization no human could manage manually. That's good for the business because you get better outcomes. 

But even setting revenue aside, we want customers to have unparalleled experiences, and personalized, relevant interactions are one part of that.

Building better chatbots

Now, back to those three buckets I mentioned earlier. That third bucket, plugging LLMs into your product, typically gets the most hype, and the biggest use case is customer service chatbots. 

You've all seen the little chat window pop up on a website. We don't have this at OpenAI right now – we invest heavily in our enterprise offering and still use human agents, with plenty of AI support behind the scenes.

But think about the power of a genuinely high-quality chatbot. Klarna has done impressive work here, and their chatbot story is worth looking up. A good chatbot does two things:

1. It saves internal resources and helps you structure customer service teams more efficiently.
2. It creates a better customer experience because people no longer wait two business days for an answer that could take two minutes.

The next wave of intelligent chatbots will look nothing like the frustrating ones we all dealt with 10 years ago. For smaller CS functions, especially startups where CSMs handle both support and core success work, this kind of technology can be a real relief.

Turning every interaction into feedback

The last thing I'm genuinely excited about is turning customer feedback into actionable insight. At most SaaS companies, feedback lives in silos. CSMs log product feedback in one system, support tracks it through tickets, and sales captures it somewhere else. The power of AI is pulling all of those inputs into one place and using it to tell you what's most impactful to build.

We want to hear from sales about what's blocking large customers from buying, and we want post-sale signals too. One thing we're exploring is parsing Gong call transcripts automatically, because so much valuable feedback gets missed unless someone logs every single thing they hear. 

We also watch engagement in our feature newsletters, where seeing that far more people clicked to learn about one feature over another tells us where the real interest lies. Marrying all of that data into one system, then using AI to think critically about it, is where everyone is heading next.

AI expands what counts as customer feedback. Every interaction becomes data you can learn from, even when it isn't formal, and that shift changes how much you can genuinely understand about the people you serve.

This article has been adapted from a talk Vanessa gave at our virtual AI for Customer Success Summit in 2024. At the time of her presentation, Vanessa was Head of Strategic Customer Success. She is currently Global Head of AI Success Engineering. 

Keep building your AI skillset

If this article has you thinking about where AI fits in your own workflow, AI for Customer Success Certified goes deeper – built by CS leaders from Cisco, Microsoft, Gainsight and Litmos, who are doing exactly what Vanessa describes.