In the current race to embed AI into every facet of business, one term is rapidly gaining traction in the customer experience (CX) space: agentic AI. While it may sound like yet another buzzword in the AI alphabet soup, there’s substance behind the surge in interest. Aaron Schroeder, director of AI and analytics solutions at TTEC Digital, believes it’s more than hype: a quiet revolution in how work gets done.
At its core, agentic AI refers to systems that can not only generate content or respond to multimodal input like text, images, or voice, but that can also make decisions and take action toward a goal, autonomously. Unlike traditional automation scripts or even generative AI chatbots, agentic AI can take a prompt or directive and then plan, execute, evaluate, and revise its actions in an ongoing loop.
“This is AI that doesn’t just talk back. It does things. It’s goal-driven. It can check whether it succeeded and keep trying until it gets there. That’s agency,” Schroeder said.
From Passive to Proactive
While generative AI excels at producing novel content and multimodal AI can take in a rich variety of inputs, agentic AI goes a step further by chaining these capabilities into action. Think of it as an AI that can run errands without constant human nudging.
“If you give it a task like ‘handle this customer inquiry,’ agentic AI doesn’t just generate a reply. It figures out what tools to use, checks its progress, and adjusts as needed,” Schroeder explains. This model mimics a classic human decision-making loop: plan, do, check, act.
Importantly, it can do so without a human in the middle, though, as Schroeder notes, many current use cases still involve human oversight. “We’re seeing the shift from augmenting humans to replacing portions of tasks entirely. That’s where this gets interesting,” he said.
Why CX Is Paying Attention
Customer experience and contact centres in particular are among the earliest adopters experimenting with agentic AI. These environments are full of repetitive, nuanced, or complex workflows that standard automation struggles to handle.
Basic requests like password resets or order cancellations are already automated. But what about rerouting a case that’s veered off script? Or managing a customer inquiry that spans multiple systems and requires an empathetic tone? That’s where agentic systems come in.
“In financial services and healthcare especially, you have situations that are too complex or sensitive for rigid automations. Agentic AI can step in, understand the context, and act more like a human agent, only faster, and in parallel across many tasks,” Schroeder said.
Examples range from self-managed returns in retail to dynamic itinerary building in travel, where the AI not only suggests options but books them. “It’s not just about replacing people. It’s about expanding what each person, or each system, can do,” he added.
Not All Smooth Sailing
Still, Schroeder is realistic about the hurdles. Legal risk and trust are the biggest friction points today. “We’re only just starting to understand how these models can ‘misbehave.’ And we don’t yet have a solid mental model for what AI failure looks like, the way we do with humans,” he said.
The other, less discussed challenge is that most companies simply haven’t invested in training their AI agents.
“We give our human agents onboarding, QA, coaching. But we’re deploying AI agents straight out of the box and expecting them to perform miracles,” Schroeder said. He believes widespread deployment depends on organisations fine-tuning models with their own data and building internal infrastructure, essentially giving AI agents their own L&D function.
“We’re seeing that when companies combine customer experience expertise with fine-tuned models built on curated internal data, things really start to work. But many haven’t gotten there yet,” he added.
The Timeline to Reality
Despite the barriers, Schroeder sees a near-term horizon where agentic AI will move from pilot projects into production environments. In fact, some already have. TTEC Digital was recently featured at Microsoft’s Build conference for its work integrating agentic systems using the open Model Context Protocol (MCP), which connects AI assistants to enterprise tools.
“We’re seeing the first real customer-facing use cases this year,” he noted. “Next year is when it gets interesting. I think by mid-2026, we’ll see agentic AI being used at scale in ways most businesses and consumers notice.”
Major financial institutions, historically risk-averse, are showing serious interest. “When those companies start piloting agentic AI for million-dollar-plus value cases, that tells you something. It means enterprise adoption is not far behind,” Schroeder said.
What Comes Next?
Asked what to watch beyond agentic systems, Schroeder doesn’t hesitate: coding agents and tools like Anthropic’s Claude, which now includes real-time prototyping features, are already disrupting workflows. But agentic AI remains his top pick for long-term impact.
“It’s the one that changes how we work. It makes multitasking scalable. It gives back mental bandwidth. And it helps bridge the gap between thinking and doing,” he said.
Agentic AI isn’t about creating flashy demos or sci-fi scenarios. It’s about quietly transforming the everyday decisions and drudgery that add up to real productivity and real change in the world of work.
In customer experience, where small efficiencies ripple outward to millions of interactions, that shift is already underway.