How to build an AI-first customer success organization

In two to three years, the Chief Customer Officers who win won't be the ones with the biggest teams; they'll be the ones with the best systems.

That shift is happening right now, whether you're designing for it or not.

I've worked with over 150 organizations, companies like Okta, ZoomInfo, Adobe, and SAP, and what I keep seeing is CS leaders standing at a fork in the road. Some are choosing to optimize: pushing CSMs from 20 accounts to 40, improving unit economics, trying to squeeze more efficiency out of what already exists. 

Others are going the opposite direction, betting everything on product-led growth and assuming the product alone will make customers successful. That works for a small number of companies. For the other 95%, it doesn't.

The third path, the one worth talking about, is building a system good enough that customers can progress and achieve success with your product in a structured, scalable way, where humans are still in the loop, but customers aren't entirely dependent on them.

That's what building an AI-first CS organization actually means.

Where most customer success orgs are today

Let's be honest about where we're starting from.

The majority of CS organizations were built around CSMs doing manual, relationship-driven work. Someone had to take care of customers, and CSMs did that, managing 20, 40, maybe 100 accounts through one-on-one meetings and personal outreach. That model served the industry well for a long time.

Then came the AI-enhanced CSM: someone with a few tools layered on top. Insights about engagement drops. AI-generated email drafts. Call summaries. Some basic workflow automation. Helpful, sure, but still fundamentally the same model with a few productivity gains bolted on.

Most companies are somewhere in those first two stages right now. The goal should be to move at least one step further to the right this year:

An AI CS organization

This is where agents are doing real work, there's an actual AI strategy mapped out by quarter, and cross-functional processes connect sales, product, and post-sale in ways that weren't possible before.

A genuinely hybrid organization

This is where I believe most of us will need to be within the next two years, managing both humans and agents. Our agents might match or exceed our CSM headcount. We’ll have a strategic AI motion, not just a plan on a slide, but a real operational advantage. And we’ll have a seat at the revenue table in a way that CS historically hasn't.

The efficiency question CS has been asking for a decade

For years, the central question in CS has been: how do we make our CSMs more efficient? How do we get someone managing 20 accounts to manage 30, then 40, then 50?

Doubling a CSM's capacity over a decade was considered a real success. But that question may no longer be the right one to ask.

The more useful question now is: how do we make our customers more autonomous? How do we build systems where customers can progress, find value, and take the next right action without always needing a CSM to guide them there?

These are two very different organizational philosophies:

It’s worth noting that this doesn't mean removing humans from the process. It means designing systems where humans intervene strategically rather than being the only mechanism through which customers can move forward.