General Catalyst backs Bolna to scale multilingual voice AI India

General Catalyst has backed a new seed round for Bolna, a Mumbai‑born startup building an orchestration layer to make multilingual voice automation reliable at enterprise scale. The founders, Maitreya Wagh and Prateek Sachan, designed the product to survive the messy realities of Indian telecoms — code‑tested systems that perform well in lab demos but fracture under real‑world load.

India still runs large swathes of commerce by phone: enterprises manage an estimated billion business calls every day, many of them multilingual and latency‑sensitive. Bolna’s thesis is simple but consequential: voice automation only helps when it is treated as infrastructure. Rather than shipping another single vendor agent, the company has built a control plane that stitches together speech‑to‑text, text‑to‑speech, LLMs, telephony and CRM integrations so enterprises can route each call to the best model, codec and network path.

The founders draw on operational experience — Maitreya from Datamuni and Bain, Prateek from Zomato and Atlassian — where they repeatedly saw prototypes break when exposed to thousands of concurrent calls, language switches mid‑sentence and intermittent network connections. Bolna now runs production systems that handle thousands of concurrent calls for customers and supports 10+ Indian vernaculars including Hindi, Tamil, Telugu, Marathi and Kannada without dropping context or mistranslating intent.

Technical novelty sits in orchestration and observability rather than a single foundation model. Bolna dynamically routes call flows to the lowest‑latency, most cost‑efficient model and keeps the stack transparent to customers. That approach addresses four common enterprise constraints — cost, latency, language coverage and auditability — and preserves human‑in‑the‑loop controls for escalations. The team reports that more than 90% of business–customer conversations can already be automated with current Voice AI capabilities, when deployed with the right telemetry and fallbacks.

From go‑to‑market to execution, Bolna emphasizes blended deployment: a Forward‑Deployed Engineer stays with a client through initial rollout to scale, reducing friction in traditional enterprise handoffs. The product is positioned as low‑code for business teams yet granular enough for engineers to tune agent behaviour, an important balance for large call centre operations that require both rapid iteration and governance.

For investors and corporates monitoring voice automation across emerging markets, Bolna’s model offers a playbook: treat voice as platform infrastructure, not a point product. Backed by General Catalyst, Bolna aims to expand beyond India into adjacent markets with similar telecom constraints and heavy voice usage — from Southeast Asia to parts of Europe where multilingual customer service creates comparable operational friction.

As enterprises push to automate routine interactions while preserving customer satisfaction, companies that can combine robust orchestration, language breadth and transparent control planes will gain an advantage. Bolna’s seed milestone signals investor belief that the hard engineering and field deployment work required to make voice AI trustworthy is itself a defensible business. If successful, the platform could unlock significant efficiency gains for small merchants and large enterprises alike, reducing manual overhead and freeing human agents for higher‑value tasks.