A Conversation with Praveen Ravela, CTO at AMC Technology: What 2026 Looks Like for CX and Agentic AI
As the year winds down, we sat down with Praveen Ravela, Chief Technology Officer at AMC Technology, to take a step back from the rapid pace of AI innovation and reflect on where customer experience is headed. In a year dominated by bold AI claims and rapid-fire pilots, his focus remains steady: what matters most isn’t just how smart these systems are, but how dependable they can become. We asked him to share his perspective on what lies ahead, the shifts CX teams should prepare for, and how leaders can separate hype from real enterprise value.
What is the most significant shift you expect to see in customer experience by 2026 as AI becomes more deeply integrated into service operations?
Praveen Ravela: By 2026, the shift won’t be about building more AI, it will be about building AI people can rely on. We already have the tools and frameworks to create agentic systems quickly. The real challenge is getting them to behave consistently when things aren’t perfect. Enterprises need predictability before they can scale, especially in customer-facing environments. AI has to handle variability in inputs without producing unexpected results. When that happens, these systems can finally move from experimental pilots to trusted parts of daily operations.
How do you see the balance evolving between human agents and AI-driven automation in 2026?
Praveen Ravela: AI will be there to support, not replace. In the next couple of years, the balance will favor hybrid workflows where AI handles repeatable, low-risk tasks and human agents focus on ambiguity, empathy, and judgment. But for that to happen, trust in the AI has to be earned. If users feel the system is inconsistent, they won’t rely on it. When the conversation shifts from “Can I trust this?” to “This helps me work faster,” that’s when the real change happens.
Which emerging AI capabilities do you believe will have the biggest impact on contact centers over the next 12–18 months?
Praveen Ravela: I expect to see a strong focus on capabilities that improve trust and visibility. Observability will be key. That means understanding how decisions are made, what signals are being used, and how AI behaves under pressure. Tools that allow for full traceability of decisions and performance under real conditions, not just in test environments, will gain traction. We’ll also see more investment in governance frameworks so teams know where autonomy is allowed and when to step in. These things will matter more than any single feature or model.
What skills or competencies will CX teams and agents need to prioritize to stay competitive in an AI-augmented environment?
Praveen Ravela: More than anything, teams need to understand how AI works in practice. This includes AI literacy, the ability to interpret results, and knowing when to trust or challenge a recommendation. Teams also need to be comfortable working within governance frameworks and adjusting workflows when new behaviors emerge. Adaptability will be key, because these systems will continue to evolve, and CX teams need to evolve with them. It’s less about technical programming and more about operational fluency.
How should organizations measure success with AI in 2026? What KPIs or business outcomes will matter most?
Praveen Ravela: Success metrics need to go beyond usage. It’s not about how many users touched the system, it’s about how consistently it delivered the right outcome. Look at things like intervention rates, reasoning trace completeness, decision volatility, and latency under real load. These show whether AI is truly ready for scale. At the end of the day, the most important questions are: Did it perform reliably? Did it support the business outcomes? And did it reduce risk or create it?
What risks or misconceptions do you think companies should be most aware of as they accelerate AI adoption next year?
Praveen Ravela: The biggest risk is assuming that a good demo means a good deployment. Many AI solutions are built as patchwork assemblies from multiple vendors. That can work fine in a proof of concept, but it often breaks down under real conditions. Things like system latency, data access gaps, and fragile integrations can create hidden risks. I also think there’s a misconception that governance slows things down. In reality, the absence of governance is what slows you down, because every rollout becomes a debate. If you want to move fast, you need clear rules that reduce friction and protect the operation.
