Desible.ai on how workflow AI is transforming India’s BFSI sector: compliance, costs, and conversion hurdles

Voice AI costs drop, but compliance and workflow challenges persist

Desible.ai co-founders Uttam Tiwari and Omkar Raikar outlined how voice AI pricing in India has fallen from Rs 7 per minute to under Rs 2 per minute, a shift driven by improved orchestration capabilities. Despite this, deploying AI in regulated sectors like banking and insurance remains complex due to compliance and workflow requirements.

The company’s platform, designed for agentic workflow orchestration, initially targeted the insurance sector. Desible.ai now handles nearly a million calls daily, reducing execution costs to below Rs 1.5–Rs 2 per minute. However, converting AI proofs of concept into production deployments remains below 35–40%, a bottleneck attributed to stringent compliance, legacy system integration, and regulatory scrutiny.

Enterprise AI adoption hinges on compliance and internal dynamics

Tiwari and Raikar emphasized that AI adoption in Indian enterprises is less about product innovation and more about navigating organisational structures. They described enterprise sales as "80% about internal dynamics and 20% about the product," where vendors must satisfy multiple stakeholders—business teams, technology departments, information security, compliance, and senior leadership—before securing approval.

Regulated industries like insurance impose rigorous information security reviews, including approximately 450 questions on data residency, compute infrastructure, and vendor risk. Desible.ai invested much of its initial funding into building compliant infrastructure to meet these requirements, rather than focusing solely on voice AI capabilities.

The co-founders noted that enterprises are shifting their evaluation metrics from technical parameters like latency to measurable outcomes such as contactability, conversion rates, and lead qualification. This reflects a broader trend toward prioritizing operational performance over technical specifications.

Global banks tighten AI agent access amid security concerns

The challenges Desible.ai addresses in India’s BFSI sector mirror global concerns about AI agents accessing customer accounts. Recent reports highlight that major U.S. banks, including Bank of America, JPMorgan Chase, and Morgan Stanley, have restricted third-party AI agents like Meta’s Muse and OpenAI’s Dots due to security risks and unclear liability frameworks. These banks have also warned customers against sharing banking credentials with AI assistants, citing risks of unauthorized transactions and data breaches.

Lazyfounder earlier reported on Bank of America’s direct advisory to customers, as well as broader industry actions to mitigate risks posed by AI agents overstepping their intended scope in financial services.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

For founders and operators in regulated industries like BFSI, Desible.ai’s experience underscores the critical role of compliance and workflow orchestration in AI adoption. The drop in voice AI costs is a tailwind, but technical prowess alone is insufficient—success depends on navigating complex organisational structures and regulatory requirements.

This story also highlights a growing divide in enterprise AI: while global banks are restricting third-party AI agents due to security risks, Indian enterprises are still grappling with deployment challenges. Startups targeting regulated sectors must build compliant infrastructure from the outset, as Desible.ai did, rather than treating it as an afterthought. The focus on operational metrics over technical specs suggests that AI vendors will need to demonstrate tangible business outcomes to win enterprise contracts.

Related on Lazyfounder

• Bank of America warns customers against sharing banking credentials with AI assistants
• US banks act against AI agents overstepping in customer accounts
• U.S. banks block AI agents like Meta’s Muse and OpenAI’s Dots over security risks
• More ai news