Five People. 100,000 Clinicians. One AI Brief.
Jerrard September has taken an unusual path to customer success. He started in the commercialization of science during his master's degree, stumbled into medical devices, then spent years shaping patient experiences across hospital networks in New Zealand, Ireland, and Australia. He took a CS lead role at an air quality company whose sensors ended up at NASA launch sites. He helped build CS at Tracksuit, a brand tracking platform that undercut the market on price and outperformed it on results. Then came EasyVet, a New Zealand success story eventually acquired by a billion-dollar US company, where Jerrard managed CS for thousands of veterinary practices across North America — a world he never expected to enter. Now he's Head of Customer Success at Carepatron, a healthcare startup serving 100,000 clinicians with a CS team of five.
The throughline across all of it is a deep instinct for scale: how to deliver meaningful experiences to large, diverse groups without losing the thread of personalization. That instinct is exactly what's driving his most ambitious AI work.
From notebooks to near-production disasters
Jerrard's entry into AI tools was social, not strategic. He runs community breakfasts for Customer Success Auckland, and at one of those meetups he met someone who had built a custom GPT trained on his own books and writing. That got Jerrard thinking about building his own "internal brain." He started with Notebook LM and ChatGPT, experimenting with how to analyze data and surface patterns from his work.
Then he moved to Claude. And that's where the real learning began.
His first ambitious project was building a lightweight internal platform that connected Claude to Pipedrive, ClickUp, and other tools in the stack — a kind of AI-powered insight layer sitting on top of the tools his team already lived in. The vision was sound. The execution had a memorable flaw: he set Claude to check for deal updates constantly, polling the system every second. Within a day, the team had burned through 20 million API credits.
"They were like, 'Wait, who is trying to check the deal update like every single second?' I inadvertently copy-pasted something and yeah, it was me." — Jerrard September, Head of Customer Success at Carepatron
It was, as he puts it, a lesson in learning the things engineers already know: polling cadence, context windows, which model you actually need for a given task. He learned them the hard way, in production, with real consequences. The experience didn't slow him down — it made him more deliberate.
What's actually working: the call prep brief
Carepatron's CS team is small by design. Five people covering 100,000 clinicians demands that every interaction be well-prepared and well-timed. Generic touchpoints don't cut it when your users range from solo practitioners running everything themselves to multi-site clinic groups with dedicated billing teams and office managers.
Jerrard's answer is a call prep brief built on top of the tools his team was already using: Pipedrive for deals, PostHog for product usage data, Intercom for support history, ClickUp for tickets, and Calendly for scheduling context. Before any CS meeting, Claude pulls from all of those systems and generates a consolidated briefing — onboarding history, usage patterns, health score, recent support conversations, and a suggested set of articles the CSM should be familiar with going in.
The most interesting element of the brief is something Jerrard calls "friction." He noticed that a large number of churned users had never contacted the support team at all — they hit a wall, gave up, and left without saying anything. So the brief now surfaces friction signals proactively: errors, call quality issues, microphone drops. If something has been going wrong in a user's experience, the CSM knows before the meeting starts.
"A large number of our users who churn never contact us. For me, friction is how we can be predictive — we know an error is happening, so we can say 'Hey, we know something may not be great. This is what you can try.'" — Jerrard September, Head of Customer Success at Carepatron
The personalization layer goes deeper than the brief. Carepatron expanded its signup flow to capture profession, team size, and role — enough to segment clinicians into distinct groups, each with their own onboarding path. A solo acupuncturist needs different things than the office manager at a multi-site psychology practice. A third-party biller has completely different needs than the clinician they support. Jerrard built those paths out deliberately, anchored in churn and expansion patterns they'd seen historically.
The goal isn't just activation — it's ensuring that by day 30, every user has had the right experience for their context.
The hard-won lessons
The most consistent theme in Jerrard's story is the cost of building first and thinking later. He's generous about sharing what that actually looked like.
The first lesson: learn how the tool works before you connect it to anything important. The 20-million-credits moment was a direct result of not understanding polling, context management, or how to structure an API integration properly. He's now built that knowledge into how he sets up every new project, and he's turned much of it into reusable skills his team can invoke themselves.
The second lesson is about setup. The decisions you make at the beginning — how you structure your prompts, how you manage context, what you ask the model to do in a single call versus spread across multiple — compound quickly. Jerrard now thinks of this the way an engineer thinks about architecture: the early calls matter more than the later ones.
"Just thinking about how you set up things at the beginning makes a world of difference." — Jerrard September, Head of Customer Success at Carepatron
The third is about team empowerment. The brief isn't a one-man tool anymore — it's something any CSM on his team can run. That shift from personal project to team infrastructure is what separates an experiment from an actual change in how work gets done.
What comes next
Jerrard comes from an implementation background, and that shapes how he thinks about the future. He's not interested in AI as a productivity hack on top of existing processes. He wants to take the whole model apart.
His bet is on full automation of the CS motion — not just the reports and briefs, but the ongoing engagement: follow-ups, nudges, communication that happens between human touchpoints. The vision is an AI that works overnight, keeps customers in the loop, and surfaces only the things that actually need a human. Fewer checks and balances, more empowerment for the team to act on what matters.
He's going deeper on Claude Code and orchestration. He thinks the onboarding and implementation process is about to look completely different, and that some roles as currently defined won't survive that shift unchanged. He's not particularly worried about that — he's building toward it.
His advice to CS teams just starting out is simple: find the moment in your day when you're stuck on a decision or staring down a task you don't want to do, and try pointing AI at it first. Start there. The sophistication follows.
Carepatron is already somewhere further down that path than most. Jerrard got there by breaking things, paying for the mistakes, and keeping his eyes on what he was actually trying to build.