Voice AI should charge for results, says Exotel’s Sachin Bhatia

Exotel cofounder and chief growth officer Sachin Bhatia expects businesses to increasingly pay voice artificial intelligence providers for results rather than calling minutes, arguing that cheaper calls alone are an inadequate measure of the technology’s value.

In an interview published by The Economic Times on October 19, Bhatia described a market split between buyers purchasing voice AI as per-minute infrastructure and those seeking a business result at a lower cost. He considers the second approach more sustainable.

Exotel supplies communications infrastructure connecting businesses with customers, alongside software for managing conversations involving human and AI agents. Bhatia distinguished that underlying infrastructure—a scale-driven business that remains important to Exotel—from the AI applications built on it.

Pricing the task, not just the call

Bhatia attributed pressure on per-minute rates to venture capital funding and competition among numerous providers. He said requests for rates of ₹2.50–₹3 a minute were common in the market, while treating discussion of ₹2 or ₹1 rates as more of a market narrative. These were his observations, not an announced Exotel tariff.

To explain outcome-based pricing, he used a hypothetical business paying ₹20,000 to generate a sales lead and asking whether AI could produce that result for ₹15,000. In that example, the relevant measure is the cost of acquiring the lead, rather than the duration of the calls involved.

He identified lead checking and payment collection as relatively low-risk tasks previously handled by sales agents that could lend themselves to result-linked pricing. Comparing the different skills and pay of insurance and institutional research salespeople, he argued that voice AI charges should similarly reflect the work performed.

For Exotel, the focus is customer service automation. Bhatia pointed to routine requests, such as order-status and account-balance enquiries, while stressing the need to bring in a human when a customer becomes frustrated.

The costs beyond a quoted rate

Bhatia cautioned that buyers need to fund the ongoing work of monitoring and improving AI agents, not merely the service’s quoted price. That includes assigning someone responsibility for the system, managing what the bot should do and regularly listening to samples of its calls.

He said Exotel is working on systems that retain context from earlier conversations, decide whether a query belongs with a human or an AI agent, and assess conversation quality. Deployment time and repeated evaluation also affect the overall cost of delivering results.

He challenged budget assumptions that 60% of conversations could be automated within three months. Automating one narrowly defined customer request might be feasible in that period, he said, but broader deployment should proceed through testing rather than an ambitious upfront target.

His suggested progression was to start with 2% automation, then move to 5% and 10% as performance is assessed, before attempting more complex requests. Those percentages were a proposed rollout approach, not reported Exotel performance figures.

Exotel explores result-linked charges

According to Bhatia, parts of Exotel’s business have already shifted from per-user charges to consumption measured in minutes or conversations. The company is also exploring outcome-linked arrangements with some design partners.

His expectation of a wider shift to outcome-based pricing remains a forecast. The examples he gave explain the commercial rationale, but do not establish measured savings or the results of those partner arrangements.