How Cars24 leverages OpenAI to expand its conversations and accelerate growth

If you run customer-facing operations at a growing company, you know how painful it is to watch potential customers slip away because no one could respond in time. This article breaks down how Cars24, a major Indian used-car marketplace, put OpenAI-powered voice and chat agents to work — handling more than one million conversation minutes every month, recovering 12% of leads that would otherwise have been lost, and rolling out agentic workflows far beyond the support team. By the end, you'll understand what they built, why it worked, and what lessons apply to your own business.

The Problem: Too Many Conversations, Not Enough Team

Cars24 operates at serious scale. Buying and selling used cars generates a constant stream of customer questions — about pricing, documentation, vehicle inspections, financing, and pickup scheduling. Traditionally, this kind of volume means either hiring large call-center teams or leaving customers waiting.

Waiting is expensive. In automotive marketplaces, a lead that goes cold often stays cold. A customer who asks about selling their car and doesn't hear back within hours may simply move to a competitor. Cars24 recognized that speed and consistency of response were directly tied to revenue, and that human agents alone couldn't cover every gap — especially across multiple languages, time zones, and channels.

The Solution: Voice and Chat Agents Built on OpenAI

Rather than building conversational AI from scratch, Cars24 leaned on OpenAI's models to power two main surfaces: voice agents that can actually hold a phone conversation, and chat agents for text-based interactions.

Together, these systems handle more than 1 million conversation minutes per month. That figure matters because it shows the agents aren't a small pilot project — they're carrying a meaningful share of real customer conversations at production scale.

Voice is the harder of the two. A phone call has no backspace key: the agent has to understand accented, noisy, sometimes incomplete speech, respond naturally in real time, and hand off smoothly to a human when the conversation goes somewhere it shouldn't. Cars24's deployment suggests the underlying models have matured enough to make voice automation practical for mainstream business use, not just demos.

The Result That Stands Out: Recovering 12% of Lost Leads

The headline number from Cars24's rollout is lead recovery. Roughly 12% of leads that would previously have been lost are now brought back into the funnel.

How does an AI agent recover a lost lead? A few ways:

• Instant follow-up. When a customer inquiry comes in outside business hours or during a rush, the agent responds immediately instead of letting the lead sit unanswered.
• Consistent persistence. Agents can politely re-engage customers who didn't respond the first time, without the fatigue or variability you'd see across a large human team.
• Coverage across channels. Someone who won't answer a call might reply to a chat message, and vice versa. Automated agents keep both doors open.

For a marketplace, a 12% lift on otherwise-dead leads compounds quickly. Those are customers who had already shown intent — the cheapest revenue the company could possibly win back.

Beyond Support: Agentic Workflows Across the Company

What makes this story bigger than a chatbot deployment is how far the agentic approach spread. Cars24 didn't confine AI agents to answering customer questions; teams across the company began building agentic workflows — systems where AI doesn't just respond but takes multi-step actions: checking records, triggering processes, escalating issues, and completing tasks end to end.

This is the real shift. A chatbot answers; an agent does. When customer-facing and internal teams can both delegate repetitive, multi-step work to AI, the technology stops being a support gimmick and becomes core operating infrastructure. Cars24's experience shows that once one team proves the value, adoption tends to spread organically.

Lessons for Your Own AI Rollout

A few takeaways worth borrowing:

1. Start where leads leak. The clearest ROI came from recovering revenue that was already slipping away, not from flashy use cases.
2. Meet customers on voice, not just chat. In markets like India, voice remains the dominant channel for many customers. Automating it unlocks far more value than text alone.
3. Design the human handoff. Agents earn trust when they know when to escalate to a person.
4. Treat agents as company-wide infrastructure. The biggest wins came after deployment spread beyond the original support use case.

Who Should Use It / Who Should Skip It

Use it if your business handles high volumes of repetitive customer conversations, loses money to slow follow-up, and serves customers across voice and chat channels. Cars24's results show the model works at serious scale.

Skip it if your conversations are low-volume, highly sensitive, or deeply nuanced — or if you're not ready to invest in testing, guardrails, and human oversight. An agent deployed carelessly can do more damage than no agent at all.