Why India needs its own agentic stack

Distyl, Sierra, Cresta. World-class US companies, deploying real AI into Fortune 500 customers, generating real customer outcomes. We have no quarrel with them — the work they’ve done has set the bar for what an agentic AI engagement can look like. We just don’t think their playbook lands cleanly in India, and we don’t think a US vendor will ever be the right answer for an Indian regulated enterprise. Here’s why.

Three structural reasons

• ›The unit economics are different. A US healthcare payor pays its nurse reviewers $30–$70/hour fully loaded. An Indian health insurer pays its adjudicators a fraction of that. The exact same touchless-rate gain produces a different ROI shape, which forces different pricing, which forces different engagement shapes. A US-priced engagement is a non-starter for an Indian buyer; an Indian-priced engagement is a non-starter for a US vendor.
• ›The regulatory stack is local. DPDP, the RBI Master Direction on IT Outsourcing, IRDAI’s cyber-security guidelines, CERT-In’s 6-hour reporting timeline, ABDM’s FHIR integration spec — these are India-specific obligations. A vendor headquartered in San Francisco can learn them, but they will always be learning them in the rear-view mirror.
• ›The buyer experience is different. The expected procurement cycle, the role of the medical director, the level of board involvement, the way SOWs are negotiated — all distinctively Indian. A vendor with a US-trained sales motion mis-times the entire engagement.

Why this isn’t a nationalist argument

This isn’t a ‘buy local’ case. It’s a fit-to-context case. Indian enterprises bought SAP and Oracle for decades because the international option was genuinely better. They bought AWS, GCP, and Azure because the local cloud option didn’t exist at scale. The right answer to ‘build vs buy local’ is whichever shape is actually fit for the constraint.

For agentic AI in regulated Indian enterprises, the constraints — DPDP’s data-fiduciary framework, RBI’s outsourcing direction, the multilingual ground reality of Indian customer interaction, the rupee-denominated economics — favour a vendor that built for them. Not because of where the vendor is incorporated, but because the engagement model and product roadmap are shaped by those constraints from day one.

We don’t need an ‘Indian alternative’ to Distyl because Distyl is bad. We need it because the constraint surface is fundamentally different.

What an India-built stack looks like in practice

Concretely, this means:

• ›Deployment in customer’s VPC as the default, not an enterprise upsell. Data-localisation is the baseline expectation, not an exception.
• ›Pricing in INR per engagement calibrated to Indian operating cost structures, not USD subscriptions calibrated to US enterprise budgets.
• ›Regulatory artefacts — exit-management plans, right-to-audit clauses, IRDAI board packs — included in the standard SOW, not bolted on for the largest customers.
• ›Multilingual not as a feature but as a default. English, Hindi, and the major regional languages from day one, including code-switched content.
• ›India-region foundation-model routing via Azure OpenAI India and Vertex Mumbai, not just ‘available on request’.
• ›Discovery calls in IST, with sales motion designed for Indian procurement cycles.

None of this is hard to do. It’s just hard to do as an afterthought. A US vendor offering India-region deployment as an option will always be slower, more expensive, and worse-fitting than a vendor that designed for it from day one.

The investor case, briefly

Indian BFSI alone is a market large enough to support multiple agentic AI vendors. So is Indian health insurance. So is Indian telecom. The total addressable spend on the human-decisioning workflows we want to automate is in the tens of thousands of crores annually, growing at double-digit percentages. The category is created; the local-fit gap is open; the regulatory tailwind is in our favour. The question isn’t whether someone builds the Indian agentic stack — it’s who.