The problem
Enterprises run millions of sales and support calls and review 3–5% of them by hand. As companies replace human agents with voice AI, the gap gets worse: nobody can say with confidence whether a bot is correct, compliant or losing conversions, or why a funnel is leaking.
What we’re building
- Voice agents for appointment booking, partner onboarding, assisted sales and driver support, with every conversation automatically checked.
- Pre-production testing: voice-to-voice and LLM-to-LLM simulation, so agents are evaluated before customers ever hear them.
- Post-production intelligence: QA on 100% of calls, funnel intelligence, A/B testing, and real-time monitoring with replay and diagnostics.
- Built for India: 13+ Indian languages, AI models hosted in India, and LLM graders that agree with human reviewers about 95% of the time.
Customer outcomes
Results our customers have reported, written up in full on prodloop.ai/customer-stories:
- Pepperfry · furniture marketplace From 3–5% of calls reviewed to 100% AI QA on every support call: manual QA time down 85%, CSAT up from 93% to 96%, first-call resolution up 10% and assisted-buying conversion up 20%.
- Vedantu · ed-tech 12% incremental revenue from smarter lead prioritization Scored every sales conversation for intent, role and blockers so agents call the right lead next: call-to-demo rate doubled and demo-to-sale rose 25%.
- Porter · logistics 20% fewer “customer not reachable” calls Analyzed 100% of calls in Hindi, Tamil and Kannada at 95%+ accuracy; root causes turned into product fixes such as platform-initiated calls and an automated waiting-time check.
- Namma Yatri · mobility Catching driver overcharging at scale Fine-tuned voice models flag drivers asking for more than the app fare across local languages; 250K+ minutes analyzed, with results feeding driver scoring and allocation.
Three pivots, each from customer evidence
We started in voice analytics, moved to evals and observability as customers began deploying voice bots, and then to agentic AI for voice operations, serving heads of sales and support directly. Each move followed where customers actually spend money and attention, not where our roadmap said they should.
Running the business
As founder and CEO I own the whole loop: product, pricing, the sales cycle, delivery and renewals. Our customers include Porter, Pepperfry, Vedantu, Snabbit, Namma Yatri and Jar. I’ve raised about $450K from Google for Startups, Microsoft for Startups, AWS Activate and an NSRCEL–Kotak grant, and lead a team of forward-deployed generalists who sit close to customers. We’re part of the Velocity–Kotak BizLabs accelerator at NSRCEL, IIM Bangalore.
What it’s teaching me about product leadership
Being accountable for revenue, not just roadmap, keeps sharpening my judgment on which problems are worth solving, and gives me first-hand depth in what makes LLM systems work in production: evals, grader calibration, multilingual quality and the cost of being wrong.