AI is changing Customer Success (CS), and now the main question is not if teams will use it, but where it can make the biggest difference.
Rob Edmondson, Ironclad’s Chief Customer Officer, believes you should begin with your existing data and let it reveal any mistaken assumptions.
On a recent episode of the [Un]Churned podcast, Rob explained how Ironclad used a churn prediction model to rethink what customer health means. This led to changes in team structure, how CSMs are measured, and how they use AI in their work.
Key Takeaways
- The usage signals you track might not be the right ones. Ironclad’s churn model showed that for their upmarket customers, daily logins did not predict renewals. Instead, cyclical engagement patterns were more important.
- Downmarket and upmarket customers need different approaches. What leads to success for SMB customers can be very different from what works for enterprise customers, so your health scoring should reflect these differences.
- Adopting AI features too early can hurt results. Customers who used AI before setting up their basic data structures showed weaker signs of renewing.
- CSM goals based on customer journey progress are more valuable than just tracking activities. At Ironclad, CSMs aim to move their accounts through set stages, not just complete QBRs.
- AI governance is still a work in progress for most CS teams, and that’s normal. Rob’s team is creating specific agents linked to OKRs instead of giving everyone access to all tools.
What the Churn Model Actually Found
Ironclad is a contract lifecycle management platform that enables customers to get contracts through legal channels, sign them, and store them in a searchable repository. Contracting is inherently cyclical, which matters a lot for how you read customer health data. Ironclad set out to build a six-month churn prediction model that could flag at-risk customers before the signals became obvious. The goal was simply to use their product usage and customer interaction data to spot customers likely to churn before it happened.
Their findings challenged some of their earlier assumptions.
For upmarket customers, daily product use was not the renewal signal they expected. In large businesses, contracting happens in cycles, not every day. When the team looked for patterns that matched how enterprise contracting really works, the signals became clear, even though they didn’t match a traditional health score.
The finding for smaller customers was just as helpful. When these customers tried to use Ironclad’s AI features too early in onboarding, their results were worse. It was more important to set up basic data structures first before moving to new features. As Rob said, “Getting some of the fundamental data structures right first, as part of their onboarding, were important before they started trying to apply the AI features to it.”
This finding directly affects how the team plans onboarding conversations and how CSMs guide customers on what to focus on first.
How the Journey Stage Framework Changed CSM Accountability
The model did more than improve the health signal. It gave the team what they needed to create a structured customer journey framework with four stages, from implementation to advanced adoption and then to what Ironclad calls the mature phase. Each stage has specific product usage benchmarks that customers must meet to move forward.
What matters most is how this framework ties into CSM evaluations. Each quarter, CSMs have targets for moving their accounts a certain percentage from one stage to the next. The focus is not on the number of calls or QBRs, but on whether customers actually progress.
“It gives a really clear mandate to a CSM of how we can measure progress,” Rob said. “We can see actual adoption metrics—are they improving or going down the path?”
CS teams have wanted to move beyond activity metrics for years. Linking CSM targets to customer journey progress is one of the best examples of this change we’ve heard on [Un]Churned. For more on how top CS teams measure CSM performance, see How Microsoft Thinks About Customer Success at Scale.
How Ironclad Is Thinking About AI Adoption Internally
Rob spoke directly about the challenge of governance. His team of about 140 people uses many tools, and they are still figuring out the balance between trying new things and setting rules.
His approach is to build agents for specific OKRs, instead of giving everyone access to all tools and hoping they find what works.
Here’s one example already in use: In pre-sales, solution engineers used to spend almost an hour creating a solution fit score and summary for each opportunity. Now, an agent does the first draft in about five minutes, and the engineer reviews and updates it before it goes into Salesforce. This is not a new use case, but it shows how you can automate a task with clear inputs and outputs, letting people focus on reviewing instead of creating from scratch.
The message he’s given his team: “Twelve months from now, if you’re a CSM or a consultant and you don’t know how to use AI tools as a daily part of your job, it’s going to be really hard to be very good at your job.”
This is not a requirement to use a specific tool. It’s a reminder that expectations are increasing, and everyone is responsible for keeping up.
Hear More From CS Leaders Reinventing Customer Success
Every week on [Un]Churned, host Josh Schachter interviews CS and revenue leaders about retention, growth, and running a modern post-sales organization. Listen to the full episode with Rob Edmondson here. To get weekly deep dives by email, subscribe to the [Un]Churned Substack.
