How Bolna AI Is Helping Enterprises Win The Voice AI Race

Enterprises are moving fast to adopt voice AI. Across ecommerce, banking, education, recruitment and customer support, companies want voice agents that can sell products, onboard customers and handle queries more efficiently.

But the biggest challenge for voice AI systems is not the chatbot or the model itself. It is the work that goes behind the scenes: getting speech recognition, language models, telephony systems, voice synthesis and workflow automation to function smoothly together at scale.

This complexity is where Bolna AI sees a clear opportunity. The Bengaluru-based startup believes the real value in voice AI will come not just from building apps but from helping businesses connect all the moving parts into one reliable system. 

Founded in 2024 by IIT Delhi alumni Maitreya Wagh and Prateek Sachan, the startup positions itself as a voice AI orchestration platform that helps enterprises launch and manage voice AI agents without building everything from scratch.

Backed by a recent $6.3 Mn seed round led by General Catalyst, Bolna AI’s pitch is simple: businesses can upload a transcript, configure an agent, buy a phone number and start AI-led calling in under 30 minutes. On the back of this speed-led approach, the startup claims to have already onboarded thousands of paying customers within two years of its inception. 

As enterprises seek voice AI systems that can be deployed quickly, Bolna AI is trying to become the layer that helps companies move from pilot to production without heavy engineering effort.

The Leap From AI Recruitment To Voice AI

When Wagh and Sachan started building Bolna AI in early 2024, the duo knew they wanted to work in voice AI. However, at the time, the category barely existed in its current form. 

“Voice AI was looking to be the next big thing. Back in 2023, there was nothing called voice AI that existed,” Wagh said, speaking with Inc42.

The founders spent their first year testing several use cases. At different points of their journey, they built an AI recruitment platform, CRM-focused products and other application-layer tools. 

Soon, a pattern emerged. Many startups were building voice applications for specific industries, yet struggled to move beyond demo environments to live production. Latency, multilingual support, telephony integrations and model selection often became major roadblocks.

“What is still stopping them from going from pure demo to POC (proof of concept) to pilot to live is the quality of the voice AI calling itself,” Wagh claims. This insight led Bolna AI to reposition itself as the orchestration layer rather than the application layer.

By mid-2025, the startup shifted its focus to building model-agnostic, voice AI agents for enterprises. Today, its product universe has expanded beyond agents to broader customer support automation, including call auditing, self-learning agentic systems and operational workflows.

On the back of this, Bolna AI today claims to have more than 2,500 paying customers on its platform, including around 100 enterprises. It also claims to see more than 2,000 signups every day, which gives it a constant flow of feedback on model performance and deployment needs.

But while much of the voice AI market is still focused on replacing call centres, Bolna AI founders see the technology becoming a layer for outcome-driven automation. Driving this confidence is the startup’s humming tech engine.

Bolna’s Model-Agnostic Approach

Unlike many startups building around a single foundational model, Bolna AI’s core philosophy is model neutrality. “Every speech-to-text, every LLM, every text-to-speech, every telephony stack, every speech-to-speech stack that you have heard of is live and available on Bolna,” Wagh claims.

It is pertinent to note that Deepgram powers a large share of the global STT workloads, while ElevenLabs and Cartesia are among the most commonly used TTS providers. OpenAI models are also predominantly used for language reasoning and orchestration.

The startup claims to integrate all these as well as other leading global and Indian AI models into its tech stack, allowing customers to choose what works best for their use case. Bolna AI also maintains fine-tuned versions of models such as Whisper and Llama. 

Therefore, Bolna’s value proposition lies in helping customers navigate an increasingly fragmented voice AI ecosystem. The startup believes its orchestration approach insulates customers from rapid model shifts. If a new speech model outperforms an incumbent provider, enterprises can switch through Bolna AI without rebuilding their entire infrastructure stack.

While this edge may work globally, the problem is more acute back home in India as several vernacular gaps remain unresolved. Wagh estimates that nearly $25,000 in monthly revenue opportunities remain inaccessible. The challenge extends to Bangla, Assamese and other regional languages, where low-latency, enterprise-grade models remain scarce.

Similarly, existing noise-cancellation systems often perform poorly in multilingual Indian environments, while long addresses, location names and pin codes create recognition issues.

These limitations are prompting Bolna AI to invest in niche machine learning capabilities of its own, particularly around vernacular speech and localisation.

Chasing The Next Growth Curve

Bolna AI’s customer base today spans sectors like ecommerce, BFSI, education, recruitment, entertainment and logistics. However, the revenue mix has evolved significantly over the past year. 

While ecommerce initially accounted for nearly half of platform usage, the founders say ecommerce, BFSI and education now collectively contribute around 80% of revenue. It counts the likes of Varun Beverages, Alibaba, GoKwik, Spinny, BiteSpeed, and Meritto, among others as its clients.

The nature of deployments is also becoming more sophisticated. Initially, many customers used voice AI for onboarding calls, cart-abandonment recovery and surveys. Today, Bolna AI claims that it is increasingly handling complex sales and customer support workflows involving multiple conversational branches, graph-based agents and tool-calling architectures.

Bolna AI’s business has grown alongside this shift. After generating its first meaningful recurring revenue in 2025, Bolna AI claims that it is currently clocking nearly $175K in monthly revenue. It operates primarily on a usage-based model, charging customers per minute of orchestration while offering forward-deployed support for larger commitments. The founders claim that the startup remains profitable and operates at gross margins above 50%.

Looking ahead, Bolna AI’s near-term focus is to deepen enterprise adoption. The voice AI platform claims that it expects to reach an ARR of $5 Mn in the coming months, aided by four to five enterprise pilots that are currently nearing production deployment. Several of these contracts are said to be worth more than $500K in annual value and span both BFSI and ecommerce sectors.

It also plans to expand its team, strengthen outbound enterprise sales and continue investing in platform capabilities and specialised ML research. All in all, Bolna AI is chasing an AWS-like playbook for voice AI. Rather than competing to build the best AI model, Bolna AI wants to become the infrastructure layer for enterprise voice AI.

As voice AI moves from experimentation to deployment, Bolna AI is betting that enterprises will need a neutral orchestration layer capable of adapting to whichever models emerge next. And in a market where new speech and language models arrive every few months, this flexibility could become as valuable as the models themselves.

[Edited by Shishir Parasher]
[Creatives by Abhyam Gusai]

The post How Bolna AI Is Helping Enterprises Win The Voice AI Race appeared first on Inc42 Media.