Why Customer Success Can't Be Automated (And What AI Can Actually Do)

In this special year-end episode of the FutureCraft GTM Podcast, hosts Ken Roden and Erin Mills sit down with Amanda Berger, Chief Customer Officer at Employ, to tackle the biggest question facing CS leaders in December 2026: What can AI actually do in customer success, and where do humans remain irreplaceable?

Amanda brings 20+ years at the intersection of data and human decision-making—from AI-powered e-commerce personalization at Rich Relevance, to human-led security at HackerOne, to now implementing AI companions for recruiters. Her journey is a masterclass in understanding where the machine ends and the human begins.

This conversation delivers hard truths about metrics, change management, and the future of CS roles—plus Amanda's controversial take that "if you don't use AI, AI will take your job."

Unpacking the Human vs. Machine Balance in Customer Success

Amanda returns with a reality check: AI doesn't understand business outcomes or motivation—humans do. She reveals how her career evolved from philosophy major studying "man versus machine" to implementing AI across radically different contexts (e-commerce, security, recruiting), giving her unique pattern recognition about what AI can genuinely do versus where it consistently fails.

The Lagging Indicator Problem: Why NRR, churn, and NPS tell you what already happened (6 months ago) instead of what you can influence. Amanda makes the case for verified outcomes, leading indicators, and real-time CSAT at decision points.

The 70% Rule for CS in Sales: Why most churn starts during implementation, not at renewal—and exactly when to bring CS into the deal to prevent it (technical win stage/vendor of choice).

Segmentation ≠ Personalization: The jumpsuit story that proves AI is still just sophisticated bucketing, even with all the advances in 2026. True personalization requires understanding context, motivation, and individual goals.

The Delegation Framework: Don't ask "what can AI do?" Ask "what parts of my job do I hate?" Delegate the tedious (formatting reports, repetitive emails, data analysis) so humans can focus on what makes them irreplaceable.

Timestamps

00:00 - Introduction and AI Updates from Ken & Erin
01:28 - Welcoming Amanda Berger: From Philosophy to Customer Success
03:58 - The Man vs. Machine Question: Where AI Ends and Humans Begin
06:30 - The Jumpsuit Story: Why AI Personalization Is Still Segmentation
09:06 - Why NRR Is a Lagging Indicator (And What to Measure Instead)
12:20 - CSAT as the Most Underrated CS Metric
17:34 - The $4M Vulnerability: House Security Analogy for Attribution
21:15 - Bringing CS Into Sales at 70% Probability (The Non-Negotiable)
25:31 - Getting Customers to Actually Tell You Their Goals
28:21 - AI Companions at Employ: The Recruiting Reality Check
32:50 - The Delegation Mindset: What Parts of Your Job Do You Hate?
36:40 - Making the Case for Humans in an AI-First World
40:15 - The Framework: When to Use Digital vs. Human Touch
43:10 - The 8-Hour Workflow Reduced to 30 Minutes (Real ROI Examples)
45:30 - By 2027: The Hardest CX Role to Hire
47:49 - Lightning Round: Summarization, Implementation, Data Themes
51:09 - Wrap-Up and Key Takeaways

Edited Transcript Introduction: Where Does the Machine End and Where Does the Human Begin?

Erin Mills: Your career reads like a roadmap of enterprise AI evolution—from AI-powered e-commerce personalization at Rich Relevance, to human-powered collective intelligence at HackerOne, and now augmented recruiting at Employ. This doesn't feel random—it feels intentional. How has this journey shaped your philosophy on where AI belongs in customer experience?

Amanda Berger: It goes back even further than that. I started my career in the late '90s in what was first called decision support, then business intelligence. All of this is really just data and how data helps humans make decisions.

What's evolved through my career is how quickly we can access data and how spoon-fed those decisions are. Back then, you had to drill around looking for a needle in a haystack. Now, does that needle just pop out at you so you can make decisions based on it?

I got bit by the data bug early on, realizing that information is abundant—and it becomes more abundant as the years go on. The way we access that information is the difference between making good business decisions and poor business decisions.

In customer success, you realize it's really just about humans helping humans be successful. That convergence of "where's the data, where's the human" has been central to my career.

The Jumpsuit Story: Why AI Personalization Is Still Just Segmentation

Ken Roden: Back in 2019, you talked about being excited for AI to become truly personal—not segment-based. Flash forward to December 2026. How close are we to actual personalization?

Amanda Berger: I don't think we're that close. I'll give you an example.

A friend suggested I ask ChatGPT whether I should buy a jumpsuit. So I sent ChatGPT a picture and my measurements. I'm 5'2".

ChatGPT's answer? "If you buy it, you should have it tailored."

That's segmentation, not personalization. "You're short, so here's an answer for short people."

Back in 2019, I was working on e-commerce personalization. If you searched for "black sweater" and I searched for "black sweater," we'd get different results—men's vs. women's. We called it personalization, but it was really segmentation.

Fast forward to now. We have exponentially more data and better models, but we're still segmenting and calling it personalization. AI makes segmentation faster and more accessible, but it's still segmentation.

Erin Mills: But did you get the jumpsuit?

Amanda Berger: (laughs) No, I did not get the jumpsuit. But maybe I will.

The Philosophy Degree That Predicted the Future

Erin Mills: You started as a philosophy major taking "man versus machine" courses. What would your college self say? And did philosophy prepare you in ways a business degree wouldn't have?

Amanda Berger: I actually love my philosophy degree because it really taught me to critically think about issues like this.

I don't think I would have known back then that I was thinking about "where does the machine end and where does the human begin"—and that this was going to have so many applicable decision points throughout my career.

What you're really learning in philosophy is logical thought process. If this happens, then this. And that's fundamentally the foundation for AI. "If you're short, you should get your outfit tailored." "If you have a customer with predictive churn indicators, you should contact that customer."

It's enabling that logical thinking at scale.

The Metrics That Actually Matter: Leading vs. Lagging Indicators

Erin Mills: You've called NRR, churn rate, and NPS "lagging indicators." That's going to ruffle boardroom feathers. Make the case—what's broken, and what should we replace it with?

Amanda Berger: By the time a customer churns or tells you they're gonna churn, it's too late. The best thing you can do is offer them a crazy discount. And when you're doing that, you've already kind of lost.

What CS teams really need to be focused on is delivering value. If you deliver value—we all have so many competing things to do—if a SaaS tool is delivering value, you're probably not going to question it.

If there's a question about value, then you start introducing lower price or competitors. And especially in enterprise, customers decide way, way before they tell you whether they're gonna pull the technology out. You usually miss the signs.

So you've gotta look at leading indicators. What are the signs?

And they're different everywhere I've gone.

I've worked for companies where if there's a lot of engagement with support, that's a sign customers really care and are trying to make the technology work—it's a good sign, churn risk is low.

Other companies I've worked at, when customers are heavily engaged with support, they're frustrated and it's not working—churn risk is high.

You've got to do the work to figure out what those churn indicators are and how they factor into leading indicators: Are they achieving verified outcomes? Are they healthy? Are there early risk warnings?

CSAT: The Most Underrated Metric

Ken Roden: You're passionate about customer satisfaction as a score because it's granular and actionable. Can you share a time where CSAT drove a change and produced a measurable business result?

Amanda Berger: I spent a lot of my career in security. And that's tough for attribution.

In e-commerce, attribution is clear: Person saw recommendations, put them in cart, bought them. In hiring, their time-to-fill is faster—pretty clear.

But in security, it's less clear.

I love this example: We all live in houses, right? None of our houses got broken into last night.

You don't go to work saying, "I had such a good night because my house didn't get broken into." You just expect that.

And when your house didn't get broken into, you don't know what to attribute that to. Was it the locked doors? Alarm system? Dog? Safe neighborhood?

That's true with security in general. You have to really think through attribution. Getting that feedback is really important.

In surveys we've done, we've gotten actionable feedback. Somebody was able to detect a vulnerability, and we later realized it could have been tied to something that would have cost $4 million to settle.

That's the kind of feedback you don't get without really digging around for it. And once you get that once, you...