Mar 26, 2025 · 34m · big-technology
Zendesk CEO: AI Customer Service Agents Are Ready For Primetime
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Zendesk CEO Tom Eggemeier joins Alex Kantrowitz to discuss the launch of the Zendesk Resolution Platform, outlining how agentic AI models, domain-specific post-training, and human-in-the-loop workflows are transforming enterprise customer service operations.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 24.1% of the talking time here. How this is scored →
Alex as informed peer 3.1 Guest teaching 2.8 Guest disagreement 0.2 Alex pushing back 1.8
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 25:10 Tom rejects the mass layoff thesis
Tom directly counters the premise of inevitable net job cuts by presenting projection data showing total interaction demand will surge 3 to 5x.
Hardest push from Alex ▶ 24:29 Alex confronts Tom on job lossesAlex refuses to accept the optimistic framing around rep empowerment, directly challenging that 80% automation will lead companies to execute layoffs.
Biggest teaching moment ▶ 7:14 Tom reveals AI bots have lower error rates than humansTom educates Alex by noting that post-trained agentic bots now exhibit a lower error rate than human contact center reps on like-for-like inquiries.
Alex holds their own ▶ 5:50 Alex analyzes deterministic vs probabilistic architectureAlex leverages his own reporting on the new Alexa Plus architecture to articulate the operational risks of deploying probabilistic LLMs in corporate environments.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| AI Agents: Separating Hype from Reality | 3 | 3 | 0 | 2 | Alex opens by pressing on what constitutes actual reality versus marketing hype in the AI agent space. Tom outlines concrete enterprise resolution rates, differentiating between realistic B2C and B2B automation targets while debunking one-day implementation claims. | |
| Evolution from Rules-Based Bots to Agentic Reasoning | 3 | 3 | 0 | 1 | Alex asks for the technical differences between earlier RPA chatbots and modern generative AI agents. Tom explains the shift from 95% rules-based decision trees to reasoning LLMs with brand personalization. | |
| Managing Probabilistic Risk and Future Bot-to-Bot Interactions | 5 | 3 | 0 | 3 | Alex demonstrates subject expertise by referencing his reporting on Alexa Plus to contrast deterministic systems with probabilistic risks in business. Tom explains how Zendesk mitigates hallucination risk using post-training across billions of historical interactions. | |
| Announcing the Zendesk Resolution Platform | 2 | 2 | 0 | 0 | Alex invites Tom to detail Zendesk's product announcement. Tom articulates their resolution-focused philosophy, contrasting outcome-based metrics against competitor per-interaction billing models. | |
| Platform Implementation, Metrics, and the Copilot Experience | 2 | 2 | 0 | 1 | Alex asks for practical implementation workflows. Tom outlines A/B testing methodologies used to reassure enterprise clients that automation will not degrade customer satisfaction. | |
| Predictive Ticket Analysis and Seamless Human-AI Handoffs | 3 | 3 | 0 | 2 | Alex probes where Zendesk gets its predictive automation estimates and how human handoffs occur. Tom details their model analysis across 5 billion annual interactions and explains VIP routing and context preservation. | |
| Transforming the Agent Role: From Creator to Editor | 2 | 2 | 0 | 1 | Alex asks about the user interface for human support reps. Tom describes the transition of customer service workers from drafting manual replies to acting as editors of AI-suggested responses. | |
| The Impact of AI Automation on Customer Service Employment | 4 | 4 | 1 | 4 | Alex delivers his sharpest challenge, asking if 80% automation will inevitably result in widespread layoffs. Tom counters by arguing total digital interaction volumes will grow 3-5x, keeping human headcount flat while elevating service quality. | |
| Klarna Case Study: Democratizing AI vs. Total Automation | 4 | 3 | 1 | 2 | Alex cites his interview with Klarna's CEO regarding full AI automation and subsequent adjustments. Tom critiques the custom DIY engineering approach versus democratizing turnkey AI for standard enterprises. |
Statements from this episode (16)
Eggemeier: AI agents resolve 60-80% of B2C customer service requests
“For customer service in particular, we are seeing AI agents really transform our, particularly our business to consumer customers. We're seeing some people get 60, 70, 80%, what we call automated resolution. So, a consumer comes into their business, they can a…”
Eggemeier: Personal bots contacting companies do not work yet, but will
“I think there's a little bit of a hype right now about you having your own personal bot contacting the company. I don't think that's working really, really well yet, but it's going to come.”
Eggemeier: Customer service bots were 90-95% rules-based four years ago
“In the past three or four years ago, you would have an AI bot, but behind the covers, it was like 95% rules-based. If X do Y, decision trees, things like that. And so people would, you know, triumph it. Hey, we got a really great AI agent. It really, or AI bot…”
Eggemeier: Zendesk deploys AI agents in days versus 6-12 months historically
“In the past, you would do a lot of customization and There are six, nine, 12 month product projects because you'd have to take all this data in. You'd have to go put all these decision trees, these rules. You'd have the bot on top of this. You know, we can get…”
Eggemeier: Zendesk trains models on database of 18B customer interactions
“We actually have eighteen billion interactions that are that are anonymized in a database with people rating customer interactions, a positive or negative, real simple, thumbs up, thumbs down. And we do a lot of post training on our models, the basic models af…”
Eggemeier: Zendesk AI agents make fewer errors than human support workers
“What we find right now is on our next generation agentic AI bot we have a lower error rate than a human being from a, you know, contact center, support center, answering like for like inquiries.”
Eggemeier: Bot-to-bot customer support interactions will arrive within 12 months
“Alexa talking to an AI agent that's, you know, powered by Zendesk, I really think that's gonna come over the next six or 12 months, and we're gonna have some really, really cool use cases where, The consumer is there using their own bot to interact with a comp…”
Eggemeier: Satisfaction decline is the biggest fear in enterprise AI adoption
“One of the biggest worries we see with our customers is I believe you can go automate X percentage of my interactions. I'm worried that customer satisfaction is going to go down and companies love the loyalty loop. They love making sure that they have happy An…”
Eggemeier: Zendesk processes nearly 5B annual interactions with 10K+ AI customers
“We do about, we process about five billion dollars, five billion tickets, or five billion interactions a year almost, and so a company will have a subset of that. We look at the data, and we say, based upon what we know, we have over 10,000 AI customers right …”
Eggemeier: 80% of customer interactions will be automated within 3-5 years
“We think about 80% will be automated within the next three to five years, leaving 20% of those interactions still to go to a human being.”
Eggemeier: Customer service jobs see 50% to 100% annual turnover
“Most of those jobs have 50 to 100% turnover a year. In a lot of companies.”
Eggemeier: Customer service interactions will surge 3x to 5x within five years
“So we have a point of view that interactions are gonna go up between three and five X. The next three to five years.”
Eggemeier: Human support headcount will remain flat despite 80% AI automation
“And so what we think is going to happen to our interactions are going to go massively up. We're going to help companies automate 80% of them, and you're going to have about the same amount of human beings that you had before in the human agent role because of …”
Eggemeier: Klarna erred by attempting to automate 100% of customer support
“Klarna got a lot of things right, but we think what they got wrong is you know, they, for a little bit, they thought they could automate everything, and where, you know, maybe forgetting a little bit that the customer's always human.”