In the Prime Venture Partners Podcast, Desible.ai founders Uttam Tiwari and Omkar Raikar discussed what is necessary to create and sell agentic AI in the Indian BFSI sector. They covered topics ranging from product engineering and regulatory compliance to enterprise sales and scaling beyond the initial use case.
The cost of voice AI is rapidly decreasing, dropping from 7 rupees per minute to less than 2 rupees. However, Uttam Tiwari and Omkar Raikar believe that the true opportunity lies not just in dialogues, but in automating workflows that operate within large enterprises.
Creating a voice agent that sounds natural is no longer the most difficult task. The much harder challenge is ensuring that such an agent functions reliably inside a bank or insurance company, including adhering to regulatory requirements, integrating with legacy systems, and achieving measurable business outcomes.
For Uttam Tiwari and Omkar Raikar, co-founders of Desible.ai, this distinction defines their company's strategy. Desible develops an orchestration platform for agentic workflows for regulated enterprises, starting with a focus on insurance.
In a conversation with Gaurav Ranjan, Senior Partner at Prime Venture Partners, the founders talked about the reasons for frequent failures of AI pilot projects when transitioning to production environments, changes in corporate procurement behavior, and how the economics of voice AI can fundamentally change the contact center industry.
Uttam and Omkar have known each other for 17 years. Their previous venture, Business Toys, was acquired in 2023. Since then, they have worked with American companies such as Newsweek, Bank of America, and Nestlé, providing software and AI consulting services.
A turning point was a logistics brokerage project that involved using AI to call truck drivers to check availability and verify information. After implementation, the client reduced its contact center from approximately 20–25 people down to just three. This experience demonstrated the potential of voice AI to transform enterprise operations.
However, while exploring the Indian market through proof-of-concept projects with 12–13 enterprises, the founders realized that automating conversations alone would not create a sufficiently differentiated business.
Uttam explained: 'Voice is a commodity. I agree, but workflows are not, because a workflow is deeply integrated into organizational processes, and that is where you get the moat.' This realization prompted Desible to focus on insurance, where customer conversations are closely linked to renewal, retention, claims, sales, and service processes.
One of the most telling discussions concerned the difference between demonstrating AI capabilities and their actual deployment in regulated organizations. Uttam estimates that the conversion rate from proof-of-concepts to production deployments among the companies he observed is less than 35–40%. He argues that a successful demonstration is only the beginning of a much more complex journey.
Insurance companies require vendors to satisfy the demands of business teams, technology departments, information security departments, compliance functions, and senior technology leadership before allowing system deployment into production.
Omkar described Indian information security audits, which include around 450 questions, including requirements related to data residency and computing infrastructure. Even established international technology providers can struggle to meet these requirements.
Consequently, Desible directed a significant portion of its initial funding toward building a compliance-ready infrastructure, rather than focusing exclusively on voice capabilities.
For Omkar, consistency is critically important, as insurance conversations leave almost no room for improvisation. He emphasized: 'You need to speak the same boring language every time, staying within compliance.'
Technical challenges go beyond compliance. Voice agents must be able to handle interruptions, understand multilingual dialogues, extract information from corporate systems, and respond appropriately without losing context.
Omkar gave an example of how an AI agent conducting a medical survey must recognize a customer interruption before hearing the full question. In another scenario, the system must identify a mentioned medication, check it against a database, and ask a relevant follow-up question. These capabilities distinguish a production-ready system from a mere conversational demo.
The pricing of voice AI has undergone sharp changes. As Gaurav noted, traditional human-staffed contact center operations can cost around 20–30 rupees per minute, whereas early voice AI solutions cost 5–7 rupees per minute. Competitive prices have since dropped to approximately 1.5–3 rupees per minute.
For startups developing voice AI products, this raises an awkward question: if the core technology becomes cheaper, how do they capture a significant share of the value they create?
Omkar countered that per-minute pricing was never a natural economic model for contact centers, which historically charged based on headcount or seats. The advent of voice AI introduced a pricing structure based on infrastructure and API consumption.
Desible is trying to move beyond this model by charging for workflow completion rather than treating every call as an isolated transaction. According to Omkar, the company processes nearly a million calls daily, and thanks to its orchestration capabilities, it has managed to reduce internal call handling costs from approximately 7–8 rupees previously used to below 1.5–2 rupees.
Nevertheless, the founders admit that corporate clients are still assessing whether AI delivers sustainable returns. Uttam shared an example of a client who did not renew a contract for three small regional contact centers after the Desible solution had been running for about four and a half months.
He warned that the overall transition remains gradual, as enterprises require sustained operational proof before reallocating budgets.
A year ago, corporate clients primarily evaluated voice AI based on metrics like latency, tone, language switching, and how natural the agent sounded. Today, Uttam explained, clients increasingly compare AI systems directly with human employees based on criteria such as availability, conversion rates, lead qualification, and accuracy of information recorded after calls.
The shift from technological novelty to measurable operational performance is also changing Desible's approach to enterprise sales. Instead of chasing every potential AI experiment, the company prioritizes workflows that already exist within the organization. Such processes have established owners, operating procedures, budgets, and key performance indicators, which simplifies demonstrating the value of automation.
Uttam cautioned: 'Your champion may not be your buyer.' A business leader enthusiastic about an AI pilot project may not have the authority to approve implementation. Information security, technology, compliance, and procurement teams can influence the final decision.
For founders selling enterprise AI, identifying these stakeholders early can be as crucial as product development itself.
Perhaps the most striking observation arose when Gaurav asked whether enterprise sales constituted 40% of the product and 60% of navigating internal organizational dynamics. Omkar suggested that in India, the latter could account for up to 80%. He argued that even a technically superior product can fail if the decision-making structures are not understood, internal support is not secured, and relationships with multiple stakeholders are not maintained.
Desible's expansion strategy illustrates this approach. At one insurance company, the company started with managing the sales funnel and then expanded to greeting calls, retention, and inbound customer service. Each successful workflow builds additional trust and reduces the effort required for subsequent integrations. Only in the area of insurance renewals has the company identified over 20 different workflows.
The founders believe that this combination of industry specialization, regulatory readiness, and integration into existing corporate processes creates a more long-term advantage than voice quality alone.
Despite high competition in voice AI, Omkar believes the market remains underexplored, especially in India, where numerous languages, dialects, background noise, and frequent language switching make conversational systems significantly more complex. He believes that companies solving these problems at scale can become global leaders.
However, Desible's long-term ambitions extend beyond call handling. The founders envision an orchestration platform that supports insurance companies and intermediaries, including third-party administrators, brokers, and health aggregators, across all their operational workflows. The ultimate measure of success will not be how convincingly the AI agent speaks, but whether it helps enterprises make better decisions, improve customer outcomes, and operate more efficiently.
Uttam's assessment of the current industry state reflects a central lesson for AI founders: 'POC is overrated, and production is underrated.'
As enterprises move from AI experimentation to demanding measurable results, the winners may not be the companies with the most impressive demonstrations, but those capable of transforming AI into reliable infrastructure for daily business operations.
