Executive Summary

• We recently wrapped up our Expert in Residence | CS Ops + AI program, an amazing collaboration with Troy Gedlinske, Senior Manager of Customer Support at Amazon/Kuiper. The goal of this collaboration was to help CS leaders develop a realistic, effective AI strategy. 
• Looking back over our journey with Troy, a few things stand out—notably the obvious benefits of investing in AI-powered assistants to expand self-service options for customers. 
• Like any good intellectual exercise, a few interesting questions emerged from the experience, including one that we talk a lot about at CDI: given the obvious ROI of proper planning, why do so many companies still struggle with it?

• Join us on February 20 to learn more. Troy will be joining Rachel Whitehorn, Luke Kempen, and Joseph Pagano of CDI for a wrap-up discussion on this fantastic program. Sign up today!

From casual conversation to amazing collaboration

Seven months ago, Troy Gedlinske of (currently of Amazon/Kuiper) joined Luke Kempen and me for a virtual chat to discuss AI and customer service. Our goal was to brief Troy about the latest trends in AI and how they might affect CS strategy.

But that’s not what happened.

Instead, Troy peeled back the curtain on the world of high-stakes, large enterprise CS strategy—offering us an inside expert’s view of this complex world.  

Thus the Expert in Residence in CS Ops + AI program was born. This amazing collaboration between Troy and CDI focused on answering an important question: how should CS leaders craft a CS Ops + AI strategy for the new era of AI? 

The output (so far) is five papers, seven blog posts, an upcoming webinar, and a workshop series that combines what we learned from Troy with CDI’s expertise in planning, design, building, and optimizing world-class Conversational AI experiences. Organized around the five pillars of CS (Quality, Finance, Scale, Flexibility, and Risk), this output infuses best practices familiar to CS leaders with a new perspective based on the latest thinking on AI tools for CS. 

In other words, something for everybody!

Preview: Four things we learned (and four questions they inspired)

In advance of next week’s webinar, I’d like to share four things we learned while working with Troy—and four questions that we, as a CAI community, need to ponder.

1. Self-service support is the foundation of modern CS—and AI-powered assistants are the self service killer apps. 

This one-two punch kept coming upon over and over again, and how could it not? 

Self-service support options are the foundation of high-quality, cost-effective, scalable, flexible, and risk-mitigated CS; this trend pre-dated AI and has a lot of momentum behind it. 

Meanwhile, AI-powered assistants are the killer app of self service: more effective and versatile than an FAQ or help doc; available 24/7/365 without breaks or unplanned interruptions; and readily extensible across multiple channels, including voice.

Our conclusion: every company should be at least experimenting with an AI assistant—provided they plan it properly and thoughtfully (more on that below).

Question to ponder: Could too much self-service “transactionalize” the customer experience, weakening brand loyalty? 

This risk is already present in self service. But if AI accelerates the effectiveness and therefore the adoption of self service, the risk could become more acute.

2. CS Ops is a specialized field that CAI practitioners must understand better.

Of course many CxDs, AI Trainers, and AI engineers understand CS Ops strategy and best practices—Luke and me included. But what we learned working with Troy was still eye opening. 

Now we’re excited to share what we’ve learned with other CAI practitioners, current and aspiring. This is important because the enterprise call center is one of the most active crucibles for AI innovation and adoption. Just scan the Fortune 500 (or CDI’s roster of clients) and try to find a company in healthcare, finance, insurance, telecommunications, or retail that doesn’t have a global, decades-old call center investment primed for AI-powered improvements.

To responsibly and effectively help these companies, we must dig deeper into the strategic and financial imperatives that have shaped the evolution of the call center ecosystem and guide the CS Ops teams who manage them today. This includes developing a better understanding of the people working in these call centers, including their incentives, career paths, and day-to-day experience.

Question to ponder: Should we do more to recruit the next generation of AI leaders from among today’s call center managers, human agents, and CS IT professionals? 

These are the people who already understand the CS Ops best practices we discussed with Troy, or at a minimum are affected by them. With many of these same people facing disruption from AI, recruiting them seems like an obvious “everybody wins” opportunity.

3. CS Risk presents a useful framework for thinking about, and pitching, AI 

Risk mitigation is an often overlooked component of many human endeavors, from choosing where to live or what partner to help build your CAI program. This is understandable: most people naturally dislike thinking about how things can go wrong—especially when it uncovers hidden assumptions about ourselves and the complex systems we take for granted.

CS Ops teams can’t afford to overlook risk. Hence thinking through CS Risk, as noted in our recent article on this topic, forced us to revisit best practices related to the other four CS pillars, and try to poke holes in them. This was a perfect capstone to our collaboration with Troy.

This exercise also highlighted ways in which the language and practice of risk management can help us think about and pitch AI. It’s easy to talk about all the goodness AI can deliver. But being able to talk about all the badness AI can help companies avoid may help us have more effective conversations with clients and colleagues whose job is to focus on reinforcing the floor, not raising the ceiling.

Question to ponder: Are we talking enough about the risk of going too slow with AI? 

It’s important to talk about AI’s many risks—from bias, reputational damage, and job market disruption to the end of human society (figuratively and literally) as we know it. And of course, FOMO is still a powerful proponent.

But one thing Troy helped us understand is how useful AI and AI-powered assistants could be in at least mitigating many of the problems CS Ops teams are trying to solve. Waiting too long to exploit AI’s benefits could be as damaging as rushing ahead.

I confess that I’m biased towards planning as the single greatest guarantor of success in pretty much everything. It’s why I’m so excited about the new planning tools and workshops we’ll be rolling out soon.

Our collaboration with Troy did little to change my mind. To the contrary, what we learned is that much of CS Ops revolves around a continuous cycle of forecasting, planning, and adapting. Without planning, there wouldn’t be many examples of high-quality, financially-sound, scalable, and flexible enterprise CS.

Yet thoughtful, proper planning is still missing in many Conversational AI programs. This represents a huge opportunity for Conversational AI practitioners and our clients to promote its merits. Just imagine the new streams of demand, investment, and innovation we can unlock by advocating for the type of planning that CS Ops leaders treat as a core capability!

Question to ponder: Why do so many companies struggle with planning? 

This is a common discussion within CDI and came up frequently in discussions with Troy. And the culprits are so numerous and diverse—from human nature to the institutional consequences of quarterly reporting and planning—that it’s easy to despair.

Yet we, as an industry, have made progress, thanks in part to tools like the CDI Standards framework. These tools, built for the new era of AI, are increasingly being married to existing tools and frameworks in software development and risk management, with good results. Our challenge now is to make such successes the norm. 

Expect a lot more conversation about this over the next few months.

If you would like to be notified of additional releases in the series along with white paper content being developed from the series in 2025, you can sign up for direct notifications here