I've been building and leading Customer Success (CS) teams for almost two decades. Big companies, small ones, some that made it, some that didn't. But no matter the company or the size of the team, I've always been solving for the same thing: how do I keep my customers, my team, and my CFO all happy at the same time?
The challenge has been the same since 2010. The difference is that now, for the first time, I actually think we have a solution.
To understand where we're headed, it helps to understand where we've been. I like to think of CS history in three distinct eras, and each one was defined by how teams tried to solve the same fundamental problem: how do you scale engagement without breaking the bank?
In the open bar era, everybody drinks, and somebody else pays. Capital was free, and if you had a problem, you solved it by spending. Throw money at it. Go big or go home.
The CS function was brand new, and teams were figuring it out in real time. Getting budget for headcount was relatively painless. When churn crept up, the answer was to hire, and for a while, that worked. More CSMs meant more attention, and more attention meant better retention. Simple math.
But the lesson that emerged from that era was a hard one: churn is a lagging indicator. By the time the number moved, you were already months past the point where you could have intervened. The teams that scaled through headcount were always reacting, never anticipating. And the solution of spending more and hiring more was available, until it wasn't. That was scaling through headcount, and you can't sustainably do that anymore.
The lesson is in knowing the difference between your lagging indicators and your leading ones. Churn tells you it's over. Time-to-value, onboarding completion rates, and early adoption signals tell you when you actually have a chance to act.
Once the industry started to pay attention to margins, we embraced segmentation. The party was still going, but you had to be on the list to get in.
The logic was simple: high-value customers get high-touch, context-driven, empathetic service. Lower-value customers get a self-service model with knowledge base articles, video libraries, group webinars, and one-to-many everything. A couple of CSMs running much higher ratios backstop the rest.
The math worked, sort of. The headcount savings could outweigh the incremental churn from the lower tier. Net-net, a margin win.
But here's the problem: the model only works once. In year two, year three, where does the next round of efficiency come from? You expand the segment that gets reduced service, or you make the service level even lower for everyone in it. Follow that path long enough, and self-service isn't really self-service. It's closer to no service. You're not managing customers anymore. You're managing the economics of not managing them.
Eventually, we recognized that even the longest-tail customer needs some interaction. You can't just hand someone a library of articles and call it a relationship. So chatbots arrived, promising scalable interaction.
The theory was sound. In practice, first-generation bots were built on decision trees. You had to anticipate what customers would ask, and the context-awareness was minimal. At their best, they were an upgrade over what came before. At their worst, they were a sophisticated phone IVR that made customers want to pound zero and demand a human.
The technology kept improving. But the core problem never went away.
Through every one of these eras, we've been stuck with the same impossible choice.
On one side: what actually works. Understanding your customer, building trust, and being a real partner. Every CSM knows this is the gold standard. The problem is, it's expensive, and it doesn't scale.
On the other side: tech touch. Scalable, cost-effective, but lacking context. It sends the same email to everyone regardless of whether the executive sponsor just left or onboarding went sideways six weeks ago.
Most teams land in an uneasy middle: human touch for the top 10%, tech touch for everyone else, and quiet acceptance that the long tail is going to churn. Pick one, lose either way.
We've been stuck with this trade-off for so long that it's started to feel like the natural order of things. An excuse, even, not to push for something better.
Here's what I believe: agentic AI makes this impossible trade-off possible to solve.
We can now deploy context-aware agents alongside our teams and directly to our customers. Agents that watch, nudge, answer, escalate, and engage. They carry the context-awareness of human touch and the scalability of technology. The trade-off isn't gone, but for the first time, it isn't impossible.
That means we've entered what I'd call the “Everyone's a VIP” era. Yes, segmentation still exists. You're going to treat a million-dollar customer differently from a twenty-five-dollar one. But the harsh disparity in the engagement model, the wide gulf between haves and have-nots in your customer base, that's no longer a structural necessity. It's a choice.
This is what we're building at OnRamp with Aero. Agents for your operations teams to build and maintain playbooks. Agents for your customer-facing teams to monitor every account and catch disengagement before it shows up in your health scores. And agents that, at your discretion and under your control, engage your customers directly.
One of our early customers, Push Operations, cut their onboarding time by 61%. Their manager of professional services told us her team will spend half as much time figuring out what's happening across their book and twice as much time on real customer relationships.
Think about what half your team's time is worth. That's not a soft number. It's real dollars you can reinvest in deeper customer partnerships, faster expansion cycles, or improved retention. Or take some off the table as cost savings. The point is, the choice is now yours to make.
The imperative has shifted. It used to be aspirational to say you wanted to provide the best service at a scalable cost. Now it's the operational standard.
That means we can no longer use the impossibility of the old trade-off as an excuse not to change. Every CS leader needs a strategy for agentic AI. Not someday. Now. The teams leading the way in 2026 are going to keep widening the gap over those who wait.
A few things to hold any vendor accountable to as you evaluate platforms:
We've spent two decades choosing between what works and what scales. For the first time, that choice is optional. The question now isn't whether agentic AI is real. The question is whether you're building your strategy for it.