Founded in 2022, Hunar.AI is building voice AI agents focused on frontline workforce management in India
The startup powers over 5 Lakh calls daily across hiring, onboarding, training and employee retention workflows for enterprises like Swiggy, Zepto, Croma and Starbucks
Instead of relying on standard STT pipelines, Hunar.AI has built a proprietary hybrid audio stack designed specifically for India’s multilingual and noisy environments
India’s AI conversation is increasingly shifting toward software agents, automation stacks and enterprise copilots. But amid the rush to automate white-collar workflows, one segment has remained relatively untouched: frontline workforce operations.
For enterprises managing thousands of gig workers, sales executives, delivery partners, retail staff or construction workers, the problem is rarely limited to recruitment alone. Attrition remains high, onboarding is fragmented, employee engagement is weak, and most workforce communication still happens manually over phone calls.
This is the opportunity that Bengaluru-based Hunar.AI is betting on. Founded in 2022 by Krishna Khandelwal and Shantanu Bhattacharyya, the startup is building conversational AI agents that automate frontline workforce management workflows across hiring, onboarding, training and retention.
“80% of the job of an HR is calling. But these are high-fidelity conversations like evaluating somebody’s skills, convincing someone to join, understanding what’s happening on the ground or training someone,” said Khandelwal, who previously served as the chief business officer (CBO) at logistics platform Locus.
The startup today powers more than 5 Lakh of these calls daily and claims to be among India’s largest voice AI companies by call volume. Its customers include Swiggy, Zepto, Aditya Birla Capital, Bajaj Finserv, Croma, Dr Lal PathLabs, 1mg and Starbucks.
The founders say the company’s origins came from identifying a structural problem inside India’s labour-heavy economy.
“Everything else gets disrupted, but this workforce doesn’t. India has a very large labour surplus, massive skill gaps and very poor workforce listening infrastructure,” Khandelwal said during a conversation with Inc42.
Before building the AI infrastructure, Hunar.AI spent nearly two years running an actual recruitment agency. The founders wanted to understand how hiring conversations happened on the ground, why frontline attrition remained high, how organisations trained employees and where communication breakdowns occurred.
During that period, the startup recorded nearly 40 Lakh to 50 Lakh minutes of workforce-related conversations. These included recruitment calls, onboarding interactions, assessment conversations and retention-related discussions. That dataset eventually became the foundation of Hunar.AI’s conversational AI stack.
According to Khandelwal, most existing voice AI deployments in India today focus on low-intelligence transactional use cases like EMI reminders or payment notifications. These calls typically last under 30 seconds and rely heavily on standard speech-to-text (STT) and text-to-speech (TTS) systems.
Hunar.AI wanted to solve a harder problem and formally pivoted deeper into voice-native AI infrastructure in 2024 after real-time audio models from companies like OpenAI started becoming commercially viable.
Subsequently it built AI agents that are designed to:
These are longer, contextual conversations where tone, pauses, interruptions and speech patterns matter. As a result, Hunar.AI claims its average call duration exceeds three minutes, significantly higher than typical Indian voice AI deployments.
One of Hunar.AI’s core differentiators lies in how it processes conversations. Most voice AI systems today operate through a standard pipeline of STT systems, and then an LLM interacting with it to convert it to text, which is then converted to speech.
According to Khandelwal, that approach works well for simple transactional workflows but struggles in India’s multilingual and noisy environments.
The challenge becomes especially visible in frontline hiring conversations happening across small towns, logistics hubs, warehouses or retail stores where accents, filler words, interruptions and background noise vary heavily. To address this, Hunar.AI built what it calls a hybrid voice architecture.
Instead of immediately converting speech into text, the startup first processes raw audio inputs through a proprietary “Dynamic Config Generator” layer. The system filters irrelevant acknowledgements, detects contextual pauses and identifies language-specific filler words before sending relevant signals into the inference engine.
The architecture also allows the system to preserve voice properties like tonality, speed, interruptions, voice modulation and conversational context. These become important in hiring or assessment conversations where intent is often reflected through delivery rather than just words.
The company has also built an “Audio Regenerative Model” to handle interruptions dynamically during conversations. Instead of restarting interactions after interruptions, the system reconstructs conversational context in real time.
Another key differentiator is multilingual adaptability. Rather than depending on a single text-to-speech provider, Hunar.AI dynamically switches between providers like ElevenLabs, Cartesia and others depending on regional language performance. The startup says this flexibility has helped it gain stronger adoption across South India, especially for Telugu, Kannada and Tamil workflows.
Under the hood, Hunar.AI currently relies on models from Google and OpenAI, while also experimenting with deploying open-source models trained on proprietary workforce conversation data.
The company claims it is storing millions of minutes of multilingual workforce interactions every month, which could eventually reduce dependence on third-party inference models over time.
Hunar.AI positions itself less as a generic voice AI infrastructure company and more as an autonomous workforce operations platform. The startup’s products are structured around specific job functions rather than generic conversational APIs.
The idea, according to the founders, is similar to how human HR functions specialise over time. “On day zero, a recruiter and telecaller might not be different. But on day 30, they become massively different,” Khandelwal said.
Then, there is the onboarding edge. The startup offers both self-serve and enterprise deployments.
Companies can create AI HR agents through Hunar.AI’s platform within minutes by defining workflows, job descriptions and knowledge bases. Enterprises can also deploy fully autonomous hiring systems where agents screen candidates, conduct evaluations and schedule interviews with minimal human intervention.
The startup additionally uses conversational agents for workforce analytics and employee engagement monitoring. These systems continuously interact with employees to identify attrition patterns, operational issues or workplace dissatisfaction signals.
Hunar.AI currently operates across six major sectors:
While quick commerce contributes the largest share of call volumes, the company claims revenue contribution remains relatively balanced across sectors. Its pricing model is also function-based rather than infrastructure-based.
Instead of charging for raw voice AI usage, enterprises pay for operational workflows such as screening, onboarding or assessments. Pricing varies by use case, with screening calls typically priced between ₹15 to ₹20, while assessments or onboarding workflows can range between ₹75 to ₹100.
Hunar.AI says it currently operates at an annual recurring revenue (ARR) of $3 Mn to $4 Mn. The startup has raised multiple funding rounds, including pre-seed and seed capital from tier-one Indian investors, though it has not publicly disclosed details yet.
The company is also in the process of closing another funding round that it plans to announce formally later.
Going forward, the startup appears focused on two major priorities. The first is deeper enterprise penetration. Hunar.AI currently competes with broader voice AI startups such as Convin and others in enterprise deployments, though Khandelwal argues the company’s differentiation comes from real-world implementation capabilities and its India-specific conversational infrastructure.
The second priority is model development. As the startup accumulates more multilingual workforce interaction data, it plans to increasingly train specialised models tailored to Indian frontline workforce communication patterns.
For now, Hunar.AI’s larger bet is straightforward. While most AI companies focus on automating software tasks, the startup believes India’s next big AI opportunity lies in managing frontline workers more efficiently.
A message from 
:
Mobavenue AI Tech Limited is an AI-Powered AdTech and Consumer Growth platform helping brands reach high-intent audiences through data-driven, scalable, and transparent marketing solutions. Learn More
Source: Inc42 - Startups
