Agentic AI in Contact Centers: Alexander Kloubek on How AI-First Models Are Rewiring Customer Experience

Agentic AI in Contact Centers: From Script-Reading Agents to Autonomous Customer Experience

For decades, the contact center has operated around a relatively simple principle: customers ask, agents respond.

Technology has steadily changed that equation. Interactive voice response, omnichannel platforms, chatbots, workforce management and conversational AI have each automated parts of the customer journey. Yet much of the underlying operating model remains fundamentally reactive. Customers still navigate fragmented channels, agents still work across multiple systems, and human intervention remains the default for far more interactions than it should.

Agentic AI changes the proposition.

Instead of merely generating a response or following a predefined workflow, an agentic system can interpret intent, reason across information, determine the appropriate action, execute that action and escalate when human judgment genuinely adds value.

That shift has profound implications for the contact center.

It could change not only how customer interactions are handled, but also how contact centers are designed, staffed, measured and continuously improved.

India sits at the heart of this transformation. As one of the world’s largest contact center markets, with enormous interaction volumes and a deeply established customer-service ecosystem, India provides an important test bed for the transition from traditional, script-driven operations to AI-augmented and increasingly autonomous customer engagement.

Next Generation Contact Center Model

NTT DATA is positioning this transformation around what it describes as a fourth-generation contact center model, bringing together intelligent omnichannel routing, conversational agentic AI, AI-powered agent desktops, real-time interaction intelligence and continuous improvement.

The bigger question is whether this represents another layer of contact center technology—or a fundamental reinvention of the operating model.

To explore that question, CXQuest speaks with Alexander Kloubek, Managing Director, Global Go to Market Leader for AI/Agentic Contact Center, NTT DATA.

NTT DATA’s contact center operations span approximately 45 years of experience, more than 32,000 FTEs and 600,000 deployed seats globally. The company handles approximately 30 million interactions annually.

Its emerging model places agentic AI at the front line of customer engagement, with human agents increasingly positioned to handle complex, sensitive and high-empathy situations.

In this conversation, Alex discusses what agentic AI means in practical contact center environments, why today’s automation approaches are reaching their limits, how humans and AI will work together, and what this transformation could mean for India’s contact center industry.

---

Defining the Agentic Shift

Q1. Agentic AI is becoming one of the most discussed developments in customer service. From an operational perspective, what makes an AI agent truly “agentic,” and how is it fundamentally different from a conventional chatbot, virtual assistant or GenAI-powered contact center tool?

AK: The fundamental difference is action. Traditional chatbots answer questions. Virtual assistants retrieve information. GenAI copilots generate or recommend content. An agentic AI system understands an objective, reasons about what needs to happen, dynamically creates a plan, interacts with multiple enterprise systems, takes authorized actions and validates whether the customer’s problem was actually resolved.

Beyond understanding and responding to intent, an agentic system takes ownership of the resolution journey, it assesses what needs to happen next, coordinates across systems and takes the required actions rather than simply providing information. It also makes service more proactive, using customer context, history and preferences to anticipate needs and recommend or execute the next best action instead of waiting for the customer to navigate a predefined process.

That changes the KPI from containment to autonomous resolution. A simple way to describe the evolution: chatbots answer, GenAI assists and agentic AI resolves.

From Automation to Autonomous Action

Q2. Traditional automation generally follows predefined rules and workflows. How does an agentic AI system change the equation when it can understand intent, reason, make decisions and take action across multiple systems?

AK: Traditional automation effectively says, “If X happens, execute Y.” Agentic AI asks, “What outcome does the customer need, and what sequence of actions will achieve it safely?” It works backward from the customer’s goal rather than forward from a predefined rule.

Consider a customer saying, “My order never arrived and I need it tomorrow.” A rules engine may create a case. An agentic system can authenticate the customer, inspect the order and logistics history, determine eligibility, check inventory, arrange an expedited replacement, update the CRM and notify the customer, potentially without human intervention. A human agent handling the same request would normally switch between multiple systems, which is cumbersome and time-consuming; the agentic system completes the entire sequence in one continuous flow, in a fraction of the time.

The objective is no longer automating individual tasks. It is automating outcomes.

The Fourth-Generation Contact Center

Q3. NTT DATA describes its approach as a fourth-generation contact center model. What are the defining characteristics of this model, and what shortcomings in today’s contact centers is it specifically designed to address?

AK: We see contact centers progressing through four generations – human-centric, voice-dominated service; omnichannel engagement and workflow automation; conversational AI, typically through chat, alongside AI-assisted humans; and now AI-first, agentic orchestration with humans in the loop.

The fourth generation turns that architecture inside out – AI becomes the first line of resolution, understanding intent, coordinating across systems and acting dynamically, while humans become an intelligent escalation layer for complexity, judgment and empathy. In a mature model, multimodal AI agents can address around 70% of all contacts. The shift matters because customer expectations keep rising while contact centers face high volumes, agent turnover and rising operational costs.

But the problem we are solving isn’t simply high labor cost. It is the fragmentation of channels, data, workflow, analytics and automation that forces customers and agents to compensate for poorly connected technology. Combined with CRM and omnichannel platforms, agentic AI carries context and historical communications and interactions consistently across voice, chat, email and social channels, moving the contact center from reactive support toward proactive, predictive and value-creating engagement.

The Five AI Layers

Q4. Your model brings together intelligent omnichannel routing, conversational agentic AI with invisible sub-agents, an AI-powered agent desktop, real-time voice mining and agent assist, and SmartCX. How do these five layers work together rather than functioning as disconnected technologies?

AK: The five layers create a closed-loop operating system for customer experience, with each layer contributing to a different stage of the resolution journey.

Intelligent orchestration identifies intent and determines the right path. Conversational AI becomes the customer-facing agent, while specialized sub-agents perform tasks such as authentication, knowledge retrieval, transactions or fulfillment. When human intervention is needed, the AI-powered “One Agent” desktop transfers context rather than merely transferring the call. Real-time voice mining then guides the human agent with prompts, knowledge and next best actions, while SmartCX analyzes up to 100% of the resulting interactions and feeds insights back into the system.

The result is a continuous loop of listening, understanding, deciding, acting, learning and improving. 

That last step is crucial. The contact center stops being a collection of separate AI tools and becomes a learning system that continuously improves its knowledge, decision-making, agent guidance and resolution capabilities.

AI as the First Line of Defense

Q5. NTT DATA is positioning agentic AI as the front line of defense, with humans stepping in when complexity or empathy requires them. What does this human-AI operating model look like in a real customer interaction?

AK: Imagine a customer calling their insurer after a major accident. The AI can authenticate the caller, understand what has happened, retrieve the policy, establish coverage, collect the structured information and set the claim in motion. But the moment the conversation needs a person, because sentiment analysis detects distress, because bodily injury is involved or because the AI reaches the limits of its authority (based on pre-defined guardrails), the interaction moves to a human. Where that line sits reflects two things: how the organization chooses to engage its customers, and the regulatory requirements that say certain interactions must be led by people.

What matters most is what happens next: the customer doesn’t start over. The human agent sees a detailed briefing of the conversation history, the customer’s context, the actions already completed, recommended next best actions and the relevant knowledge, all on a single desktop, and AI stays alongside the employee with real-time assistance rather than disappearing at handover.

Our philosophy is therefore AI-first, not AI-only.

The Human Agent’s New Role

Q6. If AI handles a growing proportion of routine interactions, how will the role of the human contact center agent evolve? Will we ultimately need fewer agents, or different and more highly skilled agents?

AK: We would need both, though the bigger story is workforce redesign. As routine transactional demand moves steadily to AI, contact centers will need fewer people for traditional tier-one work. The interactions that do reach humans, however, will be the harder ones: more complex, more emotionally charged and more consequential for the relationship.

That changes what an agent is. Tomorrow’s agent will be a problem solver, a relationship manager, an exception handler and, increasingly, a supervisor of AI, with real-time generative AI at their side. That support is already meaningful: intelligent assist tools are driving productivity gains of around 35% for human agents. Additionally, this is where AI Quality Assurance and real-time AI KPI tracking will become increasingly more important, due to the complexity of calls handled

So yes, the workforce becomes smaller in some transactional areas, but the people who remain will be more skilled, more empowered and more valuable with every interaction they handle. The measure of a great agent shifts from how quickly they close a call to how well they resolve the moments that genuinely need human judgment.

Customer Trust and Escalation

Q7. Autonomous decision-making introduces an important question of trust. How should organizations determine which decisions an AI agent can make independently and which situations must automatically trigger human intervention?

AK: Autonomy should be governed by risk, not technological capability.

The principle we use is simple: the greater the financial, regulatory, reputational or human consequence of getting a decision wrong, the more human oversight belongs in the process. Resetting a password can safely be almost fully autonomous. Waiving a small fee can sit within predefined authority limits. But denying an insurance claim, changing critical healthcare information or committing to a large financial decision deserves progressively stronger controls, and often a human sign-off.

What makes this workable is an AI authority matrix: a clear definition, journey by journey, of what an agent may recommend, what it may execute, when it must escalate and what it should never attempt, built into the guardrails from day one rather than policed after the fact.

The goal isn’t maximum autonomy. It is maximum safe autonomy; i.e. Responsible AI

Measuring 100% of Interactions

Q8. NTT DATA’s SmartCX approach evaluates 100% of interactions rather than the traditional 2–4% sample. What new insights become possible when organizations can analyze every interaction, and how can those insights translate into measurable CX improvement?

AK: Sampling tells you what happened in a handful of conversations. Analyzing every interaction tells you what is happening to your business.

Breaking Down the Technology Silos

Q9. Many enterprises still operate fragmented voice, digital and customer-service environments. How important is intelligent orchestration across channels and systems to making agentic AI genuinely useful rather than simply adding another AI layer to an existing technology stack?

AK: Because customers experience one company, even when the enterprise behind that experience runs on dozens or hundreds of systems.

An AI agent that converses brilliantly but cannot reach into CRM, billing, order management, identity, knowledge and fulfillment systems is, in the end, just a better chatbot. Agentic AI becomes genuinely transformative when it can orchestrate work across all those systems while carrying the customer’s context across voice, chat, messaging and digital channels, so no one ever has to repeat themselves.

That is why application programming interface (API) integration, data fabrics and orchestration layers sit at the heart of NTT DATA’s transformation roadmap, which moves from focused pilots through integrated, cross-channel deployment to continuous optimization. Our reference architecture connects the agentic AI layer to core platforms such as CRM, telephony, billing and knowledge management through a dedicated integration and context-passing layer.

The differentiator isn’t conversational fluency. It is enterprise execution.

The Hyperscaler and AI Ecosystem

Q10. NTT DATA works with platforms including Google Cloud, Microsoft, Amazon Connect, Genesys and NICE, while also partnering with AI-native companies such as Sierra AI. How do you decide which capabilities should be built, integrated or sourced from partners in an agentic contact center ecosystem?

AK: We don’t believe enterprises need another monolithic technology stack. The winning architecture is composable: it lets you take the strongest component for each layer, replace pieces as the market evolves and avoid tying your customer experience to any single vendor’s roadmap.

Hyperscalers and contact center as a service (CCaaS) platforms, from Microsoft, Google, AWS, Genesys, NICE, and Five9 bring tremendous foundational capabilities. AI-native companies add unique technological value in specialized areas such as Conversational AI. NTT DATA uses a blend of both, combined with proprietary solutions, while considering the client’s existing architecture to create an AI-first service model that delivers on the Board of Director and Senior Leadership goals.

Our role is to bring it all together as a complete managed service: the architecture, integration, industry intelligence, operating model, transformation capability and governance that turn those technologies into measurable business outcomes.

AK: Its purpose is to turn AI from experimentation into industrialized capability. NTT DATA established the OpenAI Center of Excellence following its global strategic collaboration with OpenAI. The CoE brings OpenAI expertise together with NTT DATA’s industry knowledge to develop services including industry-specific Smart AI Agents.

The reason it matters is that enterprises rarely fail for lack of impressive AI demonstrations. Where they struggle is the distance between a compelling prototype and secure, governed, integrated production at enterprise scale. That is a complete journey that demands orchestration and engineering expertise that a demo never tests; e.g. security reviews, integration with core systems, compliance sign-offs and an operating model that can run the solution reliably every day.

The CoE exists to close that gap, with reusable patterns, architecture, security, governance, industry use cases, engineering capability and deployment as well as operational expertise.

AK: Every enterprise deploying autonomous AI needs five things: identity, authority, observability, accountability and intervention.

In practice, that means being able to answer, at any moment, which AI agent acted, what data it accessed, what decision it made and why, what action it took and whether a human can review or reverse it. These are exactly the questions a regulator, an auditor or a customer will eventually ask.

Governance of this kind cannot be bolted on after deployment. It has to be designed into the architecture through clearly defined and aligned guardrails, spanning brand safety, with strict boundaries on sensitive subjects and “can’t do” scenarios, policy enforcement, with business rules embedded directly into agent logic, and security, with proactive handling of data privacy and compliance.

AK: The potential 40% – 65% reduction in total cost of ownership (TCO) does not come from any single lever, and it is best understood as an upper-bound transformation potential for the transactions AI directly touches, rather than a universal outcome.

The economics build layer by layer. Autonomous resolution shrinks the volume that needs human handling. Generative AI makes the remaining human work more productive. Intelligent routing cuts unnecessary transfers, automation reduces after-call work, analytics eliminates avoidable contacts and better resolution means fewer repeat interactions. Even for operations that already benefit from offshore labor arbitrage, operating costs 20%-30% lower are realistic today, with initial results visible within three months and the majority of those benefits rapidly realized between six to 12 months.

AK: India brings together scale, talent, process expertise and enormous interaction volumes. Running service operations at this scale builds a deep, practical understanding of how customer service works, and that operational knowledge matters as much as the technology when agentic AI systems have to be designed, trained and supervised well.

The sheer scale of operations makes India an extraordinary living lab for human-AI operations, where patterns proven at Indian volumes can be refined and carried to the rest of the world.

India’s opportunity is to evolve from being the world’s contact center factory into its AI-enabled customer experience innovation hub: not just running the world’s service operations but designing the intelligence behind them.

India’s Next-Generation Contact Center Workforce

Q15. With NTT DATA identifying India as one of its top revenue-generating markets and aiming to move it into the company’s top five, how do you see India’s contact center workforce evolving as AI takes over increasingly sophisticated tasks?

AK: The pyramid will change. AI will steadily absorb the repetitive tier-one work: status inquiries, simple troubleshooting, account servicing and routine transactions. Humans, in turn, will move upward into complex resolution, sales, retention, regulated interactions, AI supervision and domain-specialist roles.

This shift won’t happen overnight, and it won’t happen uniformly across the industry. As it plays out, the value a person brings will come from judgment, relationship skills and depth of domain knowledge rather than processing speed.

The opportunity for India isn’t protecting every existing task. It is moving millions of workers higher up the value chain, into roles that carry more responsibility, deeper skill and more durable value. That makes reskilling every bit as strategically important as deploying the technology itself.

From Cost Center to Experience Intelligence Engine

Q16. Could agentic AI fundamentally change the strategic role of the contact center—from a cost and service operation into an intelligence engine that continuously identifies customer needs, friction points and opportunities across the enterprise?

AK: Absolutely, and this may ultimately be the most strategic impact AI has on customer experience.

Every contact center holds millions of unsolicited customer signals: why products fail, why customers leave, which competitors are winning, where digital journeys break, what customers want next and what creates unnecessary effort. For years, most of that intelligence simply disappeared into recordings and transcripts no one had the capacity to analyze.

With AI reading every interaction, those signals can flow continuously to product, marketing, operations, sales, risk and supply chain, in near real time and in the customers’ own words.

That changes the question executives get to ask. Instead of “How efficiently did we handle 10 million contacts?”, they can ask, “Why did those 10 million contacts happen in the first place, and how do we eliminate the bad ones and monetize the valuable ones?”

Customer service stops being a cost center and becomes a real-time sensor for the enterprise.

What Enterprises Should Do Now

Q17. For a large enterprise that still operates a conventional multi-channel contact center, what should its first three steps be if it wants to begin an agentic AI transformation without attempting to replace its entire environment overnight?

AK: A conventional enterprise should start by finding the economics: studying interaction demand and identifying the high-volume, high-cost, low-complexity journeys where autonomous resolution can deliver measurable value quickly.

Second, it should define the roadmap for scale, looking across every interaction and channel at complexity, technical feasibility, the judgment involved and the impact on customer trust, including where human empathy is particularly important, and deciding deliberately which journeys belong in scope. For those that do, the foundations come first: data access, integration, security, governance, observability and clear human-escalation rules.

Third, it should prove the value and then scale: two or three journeys, honest baselines for cost, containment, resolution, customer satisfaction (CSAT) and revenue, rapid deployment, and expansion as the results come in. A well-run proof of value, with a defined success plan such as a 50% resolution target, can show what the model is capable of within roughly 90 days.

AK: I believe the term “contact center” may itself become outdated. Rather than thousands of people waiting for the phone to ring, enterprises will run networks of AI agents that resolve issues continuously across voice and digital channels, sometimes before the customer even knows a problem exists. Proactive alerts, intuitive self-service and always-on AI agents will quietly settle many issues before they ever become a call.

People will focus on a smaller but disproportionately important set of interactions, the ones involving ambiguity, emotion, negotiation, judgment, vulnerability and trust. AI will handle the volume. Humans will handle the moments that matter.

And the most advanced enterprises will stop measuring how efficiently they answer contacts. They will measure how intelligently they resolve customer needs, and how often they prevent the need for contact altogether.

Closing: The Transition

Agentic AI may ultimately prove to be more than the next technological upgrade for the contact center. Its larger significance lies in the possibility of changing the operating model itself.

The transition from scripted interactions to systems capable of understanding, reasoning and acting could fundamentally redefine the relationship between customers, technology and human agents.

Yet the success of that transition will not be determined by autonomy alone. Trust, governance, integration, workforce transformation and the ability to deliver better experiences will be equally important.

For India, the opportunity is particularly significant. Its scale, technology ecosystem and deep contact center expertise give the country a natural role in shaping the next generation of customer engagement.

Alex, thank you for sharing your perspective with CXQuest and helping us understand where agentic AI is taking the contact center—and what enterprises need to do to prepare for that future.