Bengaluru’s Desible.ai raised roughly Rs 32 crore in a seed-plus round led by Prime Venture Partners this month, with existing investor Invention Engine returning, to build what the company calls an agentic AI layer for banking, financial services, and insurance - software that doesn’t just answer questions but takes actions inside a bank’s workflows on its own.
That last part is the whole sales problem, and it’s not a technical one.
BFSI buys trust before it buys capability
Banks and insurers are structurally the most risk-averse enterprise buyers in the country, for good reason - a vendor’s mistake becomes the bank’s regulatory exposure, not just a bad customer experience. Selling them a chatbot that answers questions is hard enough. Selling them software that acts autonomously inside account workflows asks a compliance team to extend trust most vendors never get asked to earn.
That means the product demo is not what closes a BFSI deal. The audit trail is. The institutions evaluating a vendor like this want a precise answer to “what happens when the agent is wrong,” not a more impressive answer to “what can the agent do.”
Why the claimed numbers matter more here than elsewhere
Desible.ai’s public claims - more than 25 workflows running, over 10 million customer engagements a month, evaluation or use by 40-plus BFSI institutions - do more work for a BFSI vendor than the equivalent stat would for a consumer startup. A consumer audience is swayed by a growth number. A compliance officer is swayed by a volume number that implies the system has already survived scrutiny from institutions with the same risk tolerance as the one currently deciding.
The expansion path doubles as the trust-building path
The company says it started with voice and is expanding into WhatsApp, SMS, and email - channels BFSI customers already use and regulators already have frameworks for. That sequencing isn’t just a product roadmap. Every new channel it ships inside an existing, well-understood compliance framework is another data point a risk-averse buyer can check against what they already know how to evaluate, instead of being asked to approve something unfamiliar all at once.
The round buys Desible.ai more engineering time. What it has to spend on, deal by deal, is still the same thing every vendor in this category spends on first: convincing a compliance officer that an AI agent failing quietly is less likely than a human analyst doing the same.