AI as the Ultimate Leveler
Seth Jones — Jones to everyone but his family — is Senior Manager of Digital Success at Vanta, where he leads down-market customer success. He joined back when Vanta was a 20-person company, which made him one of Vanta's first CSMs and the person who had to figure out, from scratch, what customer success would actually look like there. Before Vanta, he was at Lyft, fielding the offshore tickets that escalated all the way up to a human in the United States. Before tech, he worked in nonprofits and at a civic tech company that supported political parties.
But the through-line of his career isn't any of those jobs. It's teaching. And teaching, it turns out, is exactly what makes him good at AI.
The path to AI
Jones grew up in a household full of teachers. His mom was a social worker and a teacher. His older brother, the smartest person he knew, taught him everything from how to set up the VCR to how to use a computer. Today, Jones is the IT person for his entire family.
That instinct, taking something complex and making it understandable, pulled him toward customer experience, then customer success. "I just enjoy kind of getting into technical things and figuring out how to get it to a place where people can understand and be successful," he says.
His first real AI experiments weren't personal, they were professional. As a people manager watching call recordings of CSMs who weren't performing, he hit the classic problem: too many calls, not enough hours, every team member learning differently. Some of his reports were hyper-visual. Others only learned by clicking the buttons themselves.
AI gave him a way to scale customized coaching. He started pulling thematic patterns out of calls (what specific things each CSM needed to work on) instead of giving the same generic feedback to everyone. From there, AI worked its way into the rest of his life. Today, his automation threshold sits at about thirty minutes. If a recurring task takes longer than that, he tries to automate it.
What's actually working
Jones runs the scale side of Vanta CS: high-volume, tech-touch, down-market, where you can't throw a CSM at every account.
The foundation of everything is health scoring. Vanta has rebuilt its health score five times over the years, and the static frameworks always broke as new products and features got added. So Jones's team plugged a year of historical renewal data into a tool called PecanAI and let it generate a churn prediction model from active feature usage. The result was 30 to 40 percent more accurate at predicting churn than the score they'd hand-built. That score now triggers everything: when to reach out, when to insert a human, what plays to run.
On top of that sit a small set of agents that quietly run a CSM's day before they log in. One reads the inbox and drafts replies to deeply technical security questions, pulling context from the GRC and technical support teams. That matters because Vanta's customers (often 50-person startups without a security engineer) ask CSMs questions like "is it compliant if I deploy my container this way?" — and not every CSM has a security background. Another agent combs the CSM's calendar, surfaces who they're meeting that week, what each customer has done in Vanta recently, and what to focus the call on. The output lands in Slack, not in another tool the CSM has to go check.
"That's to me the beauty of AI: it's the ultimate leveler. I could have a rep on the team who maybe is not the highest performing, but I can make you a high performing rep if I can take a lot of this reasoning out of it." — Seth Jones, Senior Manager of Digital Success at Vanta
There's also a tool called Compass that Vanta uses for democratized data analysis. AEs, CSMs, anyone curious can @-mention it in Slack, ask a question about customer data, and get back visualizations and ARR breakdowns without going through a data analyst.
The hard-won lessons
The wins came with sharp lessons. Jones tells one story about building a customer slide deck end-to-end with an AI tool. Great visualizations, clean charts, all the good stuff. He got to the very last slide before realizing the AI had hallucinated their ARR. "Whoa, our ARR isn't $3 million," he remembers thinking. He'd put hours of energy into something that quietly invented numbers.
That experience shaped how he prompts now and how he thinks about trust in AI output.
"AI in some ways, especially in the beginning, was a little bit of a sycophant. It was just like, 'yes, you're doing great, this is right.' If you didn't actually take the time to kick the tires, you very easily could just be like, 'whoa, the AI told me this is right,' and not check it." — Seth Jones, Senior Manager of Digital Success at Vanta
That's also why Vanta is quietly stepping back from one popular use case: AI-generated slide decks. The team has tried versions of it for years and the output still requires too much CSM cleanup. Their current compromise is to give CSMs the right data and a templated deck instead of trying to generate the whole thing. As Jones puts it, this is why companies like Gamma exist.
What's next
Vanta's bet for the next year isn't to hire 30 or 40 more CSMs. Nobody is. The bet is that AI agents take enough reasoning out of the day that a single CSM can hold two or three times the book they carry today, and that the job itself gets 30 to 40 percent lighter in the process.
What's different about how Vanta is going about it: AI usage is now part of performance reviews, and the whole company, not a centralized GTM engineering squad, is being trained to build. A CSM builds something that works, brings it to the GTM AI team, and they enable it from there. Jones calls this democratizing AI builds, and it's the part of his job that feels most like teaching.
His message to other post-sales leaders is unsentimental: stop being afraid that you're helping AI take you out of your job. The people who learn to build are the ones who get to shape what customer success actually becomes over the next decade. The category is being rewritten right now. The only way to put a fingerprint on it is to be in the room while it's happening.