The New CX Question: What Should AI Own?

Artificial intelligence (AI) in customer experience (CX) is no longer a future-state conversation. It is already changing how organizations think about service, support, operations, and the role of human teams. That was one of the clearest signals from this week’s Virtual Executive Roundtable with Microsoft: the conversation has moved beyond curiosity. Leaders are now wrestling with implementation, governance, and impact.

And across industries, that pressure is showing up in the same places: how to move faster without losing trust, how to create efficiency without eroding the experience, and how to separate what is genuinely transformative from what is simply new. Those were among the biggest undercurrents in this week’s Virtual Executive Roundtable, where leaders came together to talk candidly about what AI is changing within their organizations right now.

What made the conversation so valuable was not that anyone claimed to have it all figured out. It was that the discussion stayed grounded in the tension leaders are actually managing every day. What does agentic AI really mean in practice? Where is it delivering measurable value today? Where do human teams still matter most?

And perhaps most importantly, are organizations using AI to modernize the same old service model or to build something better from the ground up? The conversation made one thing clear: this moment is bigger than automation. It is a very real opportunity for CX leaders to rethink what great service looks like next.

What does “agentic” really mean?

One of the most useful parts of the conversation was the effort to separate true agentic AI from the catch-all language that’s flooding the market. Leaders repeatedly came back to one simple distinction: traditional AI reacts, while agentic AI acts. Traditional AI answers the question, and agentic AI helps complete the task. It is more conversational, more proactive, and better at taking action across systems than simply surfacing information.

That distinction matters because confusion is still slowing organizations down. Several leaders noted that many companies say they are building agentic capabilities when, in reality, they are still enhancing scripted bots. That disconnect is important. If leaders use the same word to describe radically different levels of maturity, it becomes harder to benchmark progress, align teams, and make smart investment decisions.

The industry is still early

A striking theme from the discussion was how low current adoption still is. Estimates in the room placed true agentic AI adoption somewhere below 15%, with some leaders putting it closer to 3%-5%. And yet nobody sounded relaxed about that. Quite the opposite. The tone was urgent. Leaders acknowledged that while the industry is still early, the pace of change is accelerating fast enough that waiting now could create a real competitive gap later.

That is the tension many organizations are sitting in right now. They know they do not have all the answers. They also know they do not have the luxury of standing still. If competitors have already spent the last two years learning, testing, failing, refining, and scaling, what happens to the organizations that are just beginning to ask foundational questions now?

Where leaders are seeing real value today

The conversation stayed grounded in practical use cases, and that was refreshing. Leaders discussed high-volume, repeatable work such as account updates, password resets, and status checks. They also discussed more advanced uses: proactive service interventions, automated quality auditing, and knowledge management that updates in near real time based on live interactions.

One example stood out: instead of simply telling a customer that an order is delayed, an agentic system can recognize that the customer paid for expedited shipping, predict the miss before the customer complains, and proactively issue a refund. That’s not just efficiency; that’s service with foresight.

Leaders also pushed back on a narrow definition of value. This is not only about removing cost or reducing headcount. The more mature conversation is about containment, speed, routing accuracy, reduced friction, better use of human expertise, and stronger experiences overall. In other words, if your scorecard for AI is only labor reduction, are you measuring the wrong thing?

The human-in-the-loop is a strategy

If there was one point the group returned to again and again, it was this: human oversight still matters, especially in emotionally charged moments. Leaders shared examples from industries where the issue is not a simple transaction but a life event, a safety concern, or a vulnerable moment. In those cases, the role of AI is not to replace empathy. It is to recognize when empathy is required and make the handoff smarter, faster, and more informed.

That raises an important leadership challenge. Are organizations training AI only on business rules? Or are they also teaching it context, tone, and the signals that should trigger human intervention? Because the risk is obvious: a system can be technically correct and emotionally disastrous. And in customer experience, that still counts as failure.

A few shared perspectives

• “The old AI is ‘what do you want to know?’ agentic AI is ‘what do you want to do?’”
• “Instead of AI being a helper, AI is now your accountable digital labor. It’s part of your workforce.”
• “Standard AI was reactive… agentic AI is more proactive in how we solve for the customer.”
• “Traditional AI is scripted responses. Agentic AI can go out, pull from multiple sources, and make decisions.”
• “The AI would say your order is confirmed. Agentic AI will predict it won’t arrive on time and refund the expedited shipping in real time.”
• “It’s moving from supervised learning to autonomous decision-making.”
• “Agentic AI isn’t just answering questions; it’s completing the transaction end-to-end.”

This is bigger than automation

The biggest takeaway from the discussion was not that AI is moving quickly. Leaders already know that. It is the organizations making the most meaningful progress who are asking sharper questions as they move. They are not just chasing use cases. They are thinking carefully about trust, orchestration, customer expectations, employee impact, and where human connection still matters most. That is the real work in front of CX leaders now; not simply adopting AI, but shaping how it shows up in the experience in a way that is thoughtful, responsible, and genuinely useful.

What should AI handle? What should stay human? Where can we remove friction? Where should we add reassurance? And how do we use this moment not just to do service cheaper, but to do it better?

These are exactly the kinds of candid, leader-level conversations happening inside our Virtual Executive Roundtables. If you want to hear how your peers are thinking through the opportunities and tradeoffs ahead, and contribute your own perspective to the discussion, join us at a future Virtual Executive Roundtable. It is a chance to step into a smaller, high-value conversation with fellow CX leaders navigating many of the same questions in real time.

Join our mailing list to receive event invites.