With the growing adoption of digital channels in engaging customers, the role of conversational AI is transforming from being a cost-saving mechanism to being an integral part of enterprise engagement. This is especially true in India, where language diversity, size, and government regulations create a distinct set of challenges for AI solutions.
In an exclusive conversation with Express Computer, Akhil Gupta, co-founder and CPTO of NoBroker and founder of ConvoZen, talks about how enterprises need to think about conversational AI, not as a chatbot implementation, but as a decision-support and orchestration layer that is woven into business processes.
Gupta’s approach is less product-focused and more centred around use cases, system design, and enterprise readiness, providing a practical perspective on where the value of conversational AI lies and where one needs to remain cautious.
Conversational AI begins with operational problems, not models
For many enterprises, conversational AI initiatives begin with technology selection. Gupta argues this is often the wrong starting point.
“Most meaningful implementations start with very specific operational bottlenecks,” he says. “Customer interactions are often fragmented, inconsistent, and dependent on individual agent capability. AI becomes relevant when organisations want to standardise experience without losing responsiveness.”
Common early-stage use cases include call transcription, quality audits, and interaction analytics, areas where automation improves visibility rather than replacing human judgement. Over time, these systems evolve to handle structured tasks such as query resolution, appointment scheduling, or status updates.
Crucially, Gupta emphasises that conversational AI should be viewed as infrastructure, not interface. Its value lies less in mimicking human conversation and more in connecting intent to action across enterprise systems.
Why India changes the conversational AI equation
Unlike markets where a single language dominates, India introduces complexity at every layer of conversational AI design.
“Language in India is not just about translation,” Gupta explains. “It’s about dialects, mixed languages, and context. Two people speaking the same language may express intent very differently.”
For enterprises operating at scale, this has direct implications. Generic models trained on global datasets often struggle to interpret local nuance, leading to misclassification of intent or breakdowns in conversation flow. This is where many organisations are moving towards domain-specific and language-specific models, rather than relying exclusively on large, general-purpose systems.
According to Gupta, enterprises that invest early in contextual understanding, especially for customer support, collections, and service operations, see faster adoption and lower error rates.
Agentic systems: autonomy with boundaries
One of the most discussed developments in enterprise AI is the rise of agentic systems, AI that can take actions, not just respond to prompts. Gupta cautions against viewing autonomy as an end goal.
“In enterprise environments, autonomy must be constrained,” he says. “The real value comes from orchestration, linking intent detection to predefined business actions.”
In practice, this means conversational AI systems are configured to operate within clearly defined scopes. For example, an AI agent may be authorised to raise a service request, fetch account information, or guide a customer through a process, but not to deviate beyond those boundaries. This approach addresses a key CXO concern: explainability. By limiting the action space and grounding responses in business logic, organisations retain control while benefiting from speed and consistency.
The role of conversational AI in regulated industries
Adoption patterns differ significantly across sectors. In BFSI, healthcare, and public services, conversational AI is increasingly used as a decision-support layer, not a decision-maker. “These industries have low tolerance for ambiguity,” Gupta notes. “AI can assist, summarise, flag patterns, or guide users, but final authority often remains with humans.”
Use cases gaining traction include policy clarification, process navigation, and early detection of anomalies. Full automation, Gupta believes, will remain limited to narrow scenarios for the foreseeable future.
This hybrid model, with AI handling predictable interactions while humans manage exceptions, reflects a broader trend in enterprise AI: augmentation over replacement.
Sovereign AI and the case for specialised models
India’s push towards sovereign AI infrastructure has significant implications for conversational systems. Gupta believes that large language models alone will not meet enterprise needs, particularly in multilingual and regulatory contexts.
“Specialised models built for language, domain, or function will coexist with large models,” he says. “Enterprises will increasingly stitch these together based on use case.”
He also points to growing collaboration between startups, government platforms, and research institutions to build reusable language assets. For enterprises, this could reduce dependence on black-box systems and improve alignment with local compliance requirements.
Ethics, privacy, and trust in AI-mediated conversations
As AI mediates millions of customer interactions, ethical considerations have moved from policy documents to system design. Gupta challenges the assumption that human-led interactions are inherently safer. “Humans bring variability, mood, bias, and memory. AI, if designed correctly, behaves consistently.”
That consistency, however, depends on governance. Enterprises must define what AI is allowed to do, what data it can access, and how decisions are logged and audited. Privacy safeguards and bias controls must be embedded at the architecture level, not added later.
For CXOs, this reframes ethical AI from a compliance exercise into an operational discipline.
What CXOs should prioritise next
According to Gupta, conversational AI adoption in Indian enterprises is still in its early stages. Most organisations are experimenting, not scaling. The priority for the next phase is not more intelligence but better integration, connecting conversational systems with CRM, ERP, and workflow platforms so insights translate into action.
“AI will not replace enterprise processes,” he concludes. “It will sit on top of them, making them faster and more consistent.”
The message is clear: conversational AI success will be measured not by how human it sounds, but by how effectively it reduces friction, improves decision-making, and earns trust across languages, users, and systems.