As artificial intelligence reshapes customer engagement, enterprises are moving towards more proactive, personalised and autonomous interactions. However, questions around customer trust, privacy, contextual understanding and human oversight remain central to this transformation. In this conversation with Tech Achieve Media, Ranga Jagannath, Senior Director – Growth, Agora, discusses the evolving role of AI agents in customer engagement, the importance of contextual intelligence, the responsibility of brands in AI-driven interactions, and what enterprises must consider before moving conversational AI from experimentation to large-scale deployment.
TAM: Are we moving from customer expectation and engagement to customer anticipation? AI can now predict what customers want even before they articulate it. At what point does this become intrusive, and where should enterprises draw the line?
Ranga Jagannath: Customer information has always been available to systems. What is changing is how intelligently that information is being used. I believe the critical element that enterprises must not overlook is context. A system may know a customer’s purchase history, when they last interacted with a brand and what product they bought. However, it may not necessarily understand the intent behind the customer’s current interaction.
For instance, when a customer contacts a service provider, the system should not automatically overwhelm them with information about their previous purchases if that is unrelated to the present conversation. The interaction must be guided by the customer’s immediate intent and context. Enterprises must recognise that having access to information does not necessarily mean they should use all of it. The real opportunity lies in using customer information meaningfully, rather than simply demonstrating how much the system knows. Ultimately, technology should make customer interactions more relevant and intuitive, not more intrusive or frustrating.
TAM: As AI agents become increasingly capable of handling customer conversations end to end, who ultimately owns the customer relationship: the brand, the AI agent or the technology platform powering the interaction?
Ranga Jagannath: The primary responsibility rests with the brand. As a customer, I purchase a product or service from a brand, and therefore, the brand remains accountable for the experience it delivers. An organisation cannot absolve itself of responsibility by attributing a poor customer experience to its technology provider or the underlying platform. Outsourcing technology does not mean outsourcing accountability.
Customers do not necessarily know which technology platforms or service providers operate behind the scenes. Their relationship is with the brand, and that relationship must be protected. Regardless of how sophisticated AI agents become, enterprises must retain ownership of the customer experience and ensure that the technology supporting it meets the expectations they have established with their customers.
TAM: Could contextual intelligence become the next competitive advantage beyond hyper-personalisation? How can enterprises bring together fragmented signals across conversations, devices and channels to create more meaningful customer journeys?
Ranga Jagannath: I see AI primarily as an enabler of contextual understanding. As humans, we naturally connect information across conversations. We remember what someone reported earlier, understand what has happened since, and identify factors that may be contributing to a particular problem. AI should help replicate this ability at scale.
However, the responsibility lies with product managers and technology teams to ensure that the systems they build genuinely address customer pain points. It is easy to focus on developing sophisticated interfaces, advanced features and impressive user experiences. But the fundamental question should always be whether the technology solves a real problem for the customer. AI is effective when it is provided with clear objectives, appropriate guidance and well-defined guardrails. Its performance is closely linked to the quality of the systems, processes and decisions underpinning its design. I would therefore place greater responsibility on the humans designing and deploying AI systems than on the technology itself when things go wrong. Enterprises must adopt a design-thinking approach, ensuring that technology is built around customer needs rather than allowing the availability of technology to dictate the solution.
TAM: Is AI genuinely improving customer experience, or is it simply enabling organisations to deliver mediocre experiences faster? What underlying capabilities must enterprises address to make AI agents truly intelligent?
Ranga Jagannath: The real test of an AI-driven customer experience is its ability to understand conversational nuances and respond appropriately to situations that humans handle intuitively. Consider a conversation in which one person interrupts another. A human typically pauses, acknowledges the interruption and then resumes the conversation. AI systems must develop similar capabilities to manage interruptions naturally.
Another important consideration is noise management. In a real-world environment, multiple conversations may be taking place simultaneously. Humans can generally distinguish the primary conversation from background noise and focus on the person they are interacting with. AI systems must demonstrate comparable capabilities, particularly in environments involving multiple speakers, interruptions and competing audio inputs.
While AI is increasingly capable of handling one-to-one conversations, these more complex scenarios continue to present challenges. Technology teams must therefore evaluate whether their systems can handle real-world conversational conditions rather than assuming that performance in controlled environments will translate directly into production. The objective should be to build systems that understand the nuances of human communication, not merely systems that can respond to predefined inputs.
TAM: We are moving towards a scenario where customers may have their own AI shopping agents interacting with AI agents deployed by brands. Could AI-to-AI engagement become a reality, and what happens when customers become as intelligent as the systems they interact with?
Ranga Jagannath: I believe this presents an interesting but complex scenario. If two AI agents interact without adequate training or clearly defined parameters, the outcomes could be unpredictable. At the same time, making AI systems increasingly capable of understanding and performing tasks independently raises broader questions about the evolving role of humans.
Honestly, I do not believe we have definitive answers yet. There have been instances of multiple agents working together to solve complex problems, producing outcomes that were not necessarily anticipated. There will undoubtedly be use cases where AI-to-AI interactions can deliver significant value. However, enterprises must approach such deployments with caution. One possible approach is to establish clearly defined thresholds for autonomous interactions. For instance, agents may be permitted to handle transactions within specified limits, while interactions involving higher-value transactions or greater complexity could require human intervention.
The key is to balance autonomy with oversight. As these technologies evolve, enterprises will need to establish appropriate boundaries based on the nature, complexity and potential consequences of the interaction. I would describe my outlook as cautiously optimistic. The opportunities are significant, but so is the need for responsible deployment.
TAM: From Agora’s vantage point, what should enterprises consider when building conversational AI platforms, particularly as they move from experimentation to production?
Ranga Jagannath: When building conversational AI platforms, experimentation with readily available technologies and off-the-shelf solutions is perfectly reasonable. Organisations can use these tools to develop initial applications, test ideas and validate their approach. However, the considerations change significantly when moving towards large-scale deployment.
Enterprises must assess whether the technology they are building can support production requirements and deliver conversations that closely resemble natural human interactions. It is important to evaluate the available technology options, including platforms such as Agora, based on their ability to meet the organisation’s requirements. A proof of concept and a production-grade system are fundamentally different. What works effectively in a controlled demonstration may not necessarily deliver the same performance under real-world operating conditions. Enterprises must therefore exercise sound judgement when selecting their technology stack and designing systems intended for scale.
TAM: Many enterprises are experimenting with conversational AI, but scaling these initiatives into production remains a challenge. What differentiates organisations that successfully achieve large-scale deployment from those that remain stuck in the pilot phase?
Ranga Jagannath: The starting point is understanding the problem an organisation intends to solve and the scale at which it needs to solve it. Enterprises must establish realistic operational requirements. For conversational AI, this could include the expected number of calls per day, average call duration, number of participants in each conversation and the overall capacity required.
These assessments help determine the infrastructure needed to support the application, including the hardware, network capabilities and technology stack. A small proof of concept may perform perfectly well on a laptop or limited infrastructure because the volume of interactions is minimal. However, production-grade systems require a much more comprehensive evaluation. Organisations must assess whether their infrastructure can support the anticipated workload, with sufficient capacity to accommodate growth. They must also evaluate the technology provider’s capabilities, the maturity of its technology stack, its financial standing and its experience in delivering scalable solutions.
Beyond the technical and financial considerations, the quality of the customer interaction itself remains important. Enterprises should not overlook the experience of the person on the other side of the conversation in their efforts to reduce costs or automate workflows. Ultimately, successful deployment requires a combination of clearly defined business objectives, realistic capacity planning, appropriate technology infrastructure, reliable partners and a strong focus on customer experience.
