AI agents make CX orchestration more important

More capable AI can make customer experience harder to coordinate. Enterprises are adding AI agents, voice AI, and automation across messaging, voice, and digital channels faster than the architecture beneath them is changing. Gaurav Anand, global head of the Customer Interaction Suite at Tata Communications, frames the problem as “Orchestration is the new challenge for CX in the age of AI agents” and says: “In the rush to deploy AI, organizations have largely bolted conversational AI onto legacy systems.” Anand says many enterprises now use digital tools, while “very few” have genuinely integrated, scaled platforms able to orchestrate them seamlessly.

More AI can create a bigger CX coordination problem

The legacy architecture matters because traditional CX systems were built around linear routing led by people. AI changes that operating model because autonomous systems, data lakes, applications, and human workers can all participate in one customer interaction and need information from one another in real time. Adding conversational AI while keeping the underlying architecture leaves each participant working through systems with fragmented data and decisions.

That fragmentation becomes visible when human agents receive a customer from an AI system and have to reconstruct what happened. They may search disconnected tools simply to discover what the AI already told the customer, increasing cognitive load and losing useful context during the handoff. Across channels, the same failure creates latency, inconsistent journeys, and repeated interactions, which can affect brand trust and customer loyalty.

Voice AI can reproduce the same structural problem when it sits in front of an old contact-center system. The customer may meet an AI agent first, while the journey still follows the deterministic phone menus that voice AI was intended to replace. Anand argues that AI’s useful advantages are scale, speed, and orchestration, so retaining old routing assumptions limits what a deployment can change.

The constraint is coordination

Once multiple AI systems can perform useful work, increasing their individual capability leaves the handoff problem unresolved. Anand puts the shift this way: “Today’s operational complexity is no longer about adding more intelligence. It is about coordinating the existing intelligence across the enterprise, so the enterprise customer never feels the friction of those internal silos.” His proposed requirement is a shared context layer through which AI systems, applications, and people work from the same understanding of both the customer and the business.

That shared layer makes orchestration a broader architectural objective than task automation. “Automation solves individual tasks, whereas orchestration connects them into end-to-end outcomes,” Anand says. He defines the next step as context-aware orchestration, in which AI agents, applications, and people share knowledge of customers, business processes, and business intent instead of acting from separate system records.

The need for orchestration grows as an enterprise accumulates bots, agents, and other AI tools. Anand says management complexity rises exponentially as those systems multiply, making collaboration, escalation, and intelligent handoffs more important as sources of competitive advantage. In his view, CX design should therefore judge whether the whole customer outcome remains coherent as work passes among systems and people.

Shared context needs shared meaning

Coordinating that outcome requires more than moving data between applications because each participant must interpret the information within the same business context. A customer identity may exist in one system, a transaction in another, a conversation elsewhere, and the relevant policy or journey state somewhere else again. Anand’s proposed context layer connects customer identities, interactions, transactions, policies, journeys, and operational systems so the shared understanding survives application boundaries.

Anand’s mechanism for establishing that understanding starts with an enterprise ontology. An ontology is a shared vocabulary and set of business concepts that gives systems common meaning for the information they exchange. Here it spans customer data, products, policies, standard operating procedures (SOPs), transactions, and workflows, allowing otherwise disconnected platforms to represent the enterprise in compatible terms.

That common meaning supports context graphs, which connect relationships among customers, interactions, products, policies, decisions, and outcomes across organizational silos. The graph can give an AI agent the relationships needed to understand how individual records belong to the current interaction. Anand argues that AI systems and human workers can then make more accurate decisions, perform smoother handoffs, and give customers a more consistent experience because each participant starts from the same contextual basis.

Those relationships must also persist as the customer moves through a journey. A context-driven architecture continuously connects identity, conversations, transactions, and operational information so each touchpoint can inherit the state established earlier. When a customer moves among voice, WhatsApp, chat, email, and CRM workflows, identity, intent, and AI-derived insight move with the interaction instead of remaining inside the application that first captured them.

Persistence makes the context layer a live operational dependency because customer state can change during the interaction itself. Customer intent can shift during a conversation; an AI system can reach a new decision; a transaction can alter the customer’s status; and a policy can determine the next permitted action. Conversation history, enterprise data, intent, and AI decisions consequently must remain synchronized as work passes between channels and systems.

Tata Communications uses its Interaction Fabric as its product-level illustration of this architecture. The company describes Interaction Fabric as an orchestration layer combining contact center, messaging, collaboration, AI, and customer data while coordinating AI agents, channels, and enterprise systems in real time. Tata Communications has a commercial interest in enterprises adopting this model because Interaction Fabric is its product, and the company positions the product around making AI a connective layer among customers, employees, and enterprise systems while retaining context across their interactions.

The proposed architecture consequently asks integration to provide common interpretation as well as technical connectivity. APIs can make information reachable, while an ontology supplies common meaning, context graphs preserve relevant relationships, and continuous propagation keeps the current state available where a person or AI system must make a decision. Together, those mechanisms are intended to keep separate participants aligned on the customer’s current situation.

Timeliness then becomes part of the same CX architecture because shared state loses operational value when updates arrive too late. Anand calls the problem created by legacy networks unable to handle modern data frequency “data gravity”: customer state and AI decisions can lag as users move among channels, producing latency and inconsistent journeys. A technically connected handoff can still produce a poor customer experience when the receiving system acts before the relevant context reaches it.

For Anand, timely propagation makes synchronous interaction an infrastructure requirement. “The underlying network needs to be engineered to be as agile as the AI systems running on top of it,” he says. “Interactions stay synchronous and technology itself becomes invisible, leaving only an experience that feels effortless.” Under this model, systems need common meaning, persistent state, and sufficiently timely synchronization so each participant can act on the same situation.

Orchestration changes where humans belong in the journey

Once AI and people share that situation, enterprises can place human judgment where the interaction requires it. Shared visibility starts with the agent experience because information gathered by an AI or human in one interaction remains available in the next regardless of the channel or system involved. That continuity reduces the reconstruction work that arises when agents inherit conversations from disconnected automation.

With that continuity in place, agent-facing AI can operate inside the common workflow. Automated call summaries give the next worker the relevant conversation state, real-time sentiment analysis exposes changes in the customer’s response, and AI assistance can surface actionable information and suggested next steps while the interaction is underway. These functions affect what an agent can decide at the moment of handoff because the worker receives useful state inside the active interaction.

The division of work can then follow the nature of the interaction. AI can handle routine, high-volume requests such as password resets, delivery tracking, and account updates. Human agents can focus on situations where judgment, empathy, and careful communication affect the customer outcome, with orchestration deciding when that transition should occur.

A fraudulent transaction makes that boundary concrete. “If a customer is facing a sudden crisis like a fraudulent transaction, the AI can instantly block the card, but it cannot provide the emotional comfort and delicate communication needed in that moment of panic,” Anand says. In his scenario, AI executes the urgent technical action, real-time sentiment analysis identifies distress, and orchestration routes the interaction to a human expert.

That sequence makes human participation a planned part of the workflow. Anand describes the answer as “intelligent orchestration, rather than a choice between systems.” The business goal is to capture AI’s efficiency while preserving the judgment and communication that support brand trust and loyalty when an interaction becomes sensitive.

The implementation problem spans architecture, network, and organization

Building that operating model requires enterprises to move beyond fragmented AI experiments. Anand’s first proposed step is to consolidate data and fragmented point solutions on a unified, cloud-first platform, reducing the number of separate places from which customer state must be reconstructed. Communication APIs then need to sit within the enterprise core so business functions can draw on common customer context as interactions move through the organization.

With basic integration established, a shared ontology and context graphs can create common meaning across CX, operations, sales, service, and AI systems. The resulting contextual architecture gives those functions a consistent basis for interpreting customer state and business intent. The network then has to keep that state synchronized at the speed required by real-time interactions.

The organizational boundary follows from those technical dependencies because CX teams design customer journeys while IT teams control much of the infrastructure and integration beneath them. “IT and CX teams need to work more collaboratively,” Anand says. A journey that crosses AI, communications infrastructure, enterprise data, policies, and human service depends on those teams aligning their operating assumptions as they modernize the components.

That dependency puts the modernization work at three connected layers: software and data need shared business meaning, networks need to preserve live synchronization, and teams need to align systems around the same customer outcome. Each layer depends on the others. A context graph cannot produce a smooth real-time handoff when infrastructure introduces material lag, while fast infrastructure cannot reconcile incompatible interpretations of customer and policy data.

Beyond individual enterprise architecture, Anand points to consolidation among contact-center providers as a market signal, saying established vendors are acquiring AI-native companies to close capability gaps and broaden their CX offerings. Tata Communications, which competes commercially in the CX and communications market and benefits from demand for broader orchestration platforms, interprets that direction as evidence that enterprises increasingly want an intelligence layer capable of coordinating AI, people, data, and workflows across the business.

More autonomous CX raises the coordination requirement

More autonomous agents increase the need for common context because software will increasingly make and execute decisions without waiting for a person to reconcile conflicting information. Anand expects CX to move from reactive service toward what he calls the “three Ps”: proactive, predictive, and personalized engagement. That operating model requires real-time intelligence and persistent enterprise context to follow customers, employees, and AI agents as they move through interactions.

The timing of CX decisions changes with that model as well. Enterprises can use live context to shape an interaction while it is happening instead of relying on analysis of a completed conversation to improve a later one. Anand sees greater autonomy and seamless orchestration across touchpoints developing together because autonomous action becomes more useful when each agent has current knowledge of the journey and the enterprise rules governing it.

That direction leads Anand to describe “simplification” as the defining direction: data, infrastructure, and operating models align around customer outcomes as AI capabilities develop. He expects AI-powered agents and agent-to-agent interactions to move from assisting people toward independently managing and resolving interactions. Human agents would work alongside those systems with real-time conversational intelligence and next-best-action recommendations available during customer engagements.

Anand calls the combined operating model “Total Experience,” bringing customer, employee, and AI-driven experiences into one design. Tata Communications says it is building toward that model through Voice AI, AI Workers, and its Total Experience Hub. These are Tata Communications products, so the company has a commercial stake in the broader architectural approach it advocates through them.

That proposed model ultimately changes when the enterprise intervenes in a customer journey. “Ultimately, customer engagement will evolve from being reactive to predictive and increasingly generative,” Anand says. “Enterprises won’t just be responding to needs, but actively shaping and improving customer journeys in real time.” As systems gain greater autonomy, persistent context gives those independent systems a coordinated view of the customer, enterprise policy, and one another while they act.

Key executive takeaways

• Make orchestration the CX priority: As AI agents, human workers, applications, and channels multiply, fragmented systems create inconsistent handoffs and customer journeys. CX and IT teams can evaluate AI investments by how well they coordinate end-to-end outcomes across these components.
• Build a shared context layer: AI agents and employees need a common view of customer identity, intent, interactions, policies, and business processes. Enterprise architects can use shared ontologies, context graphs, and persistent state to preserve that understanding across systems and channels.
• Preserve context in real time: Customer state and AI decisions lose operational value when network or integration delays leave receiving systems with outdated information. Infrastructure teams can design for real-time synchronization so context follows customers across voice, messaging, email, CRM, and other workflows.
• Design human handoffs into AI workflows: AI can execute routine and time-sensitive tasks while human agents handle interactions requiring judgment, empathy, or careful communication. CX teams can use shared context and signals such as sentiment to determine when and how those handoffs occur.
• Modernize architecture and operating models together: Unified platforms, communication APIs, shared business semantics, responsive networks, and closer CX-IT coordination are interdependent parts of orchestration. Technology owners can align modernization decisions around complete customer journeys rather than isolated AI deployments.
• Prepare for more autonomous CX: Proactive, predictive, and personalized customer engagement increases the importance of persistent enterprise context as AI agents make more decisions independently. Enterprises building agentic CX can establish orchestration and governance foundations before autonomy expands across customer journeys.