Currently, Bhatia said the voice AI industry serves buyers who either treat the technology as infrastructure billed by the minute or want an existing business outcome delivered at a lower cost, with the latter approach offering a more durable model.
“I don’t think voice AI is an infrastructure business. People are reading it that way, but I think it will stabilise to be an outcome-based business very soon,” Bhatia told ET AI.
Exotel provides communications infrastructure that connects businesses and customers, as well as software that manages conversations handled by human and AI agents.
From calling minutes to business results
Bhatia linked pressure on the per-minute pricing model to venture capital funding and a crowded market, where providers compete to offer increasingly cheaper services.
“We are finding Rs.2.50 to Rs 3 per minute to be a common ask in the market. There is also talk of Rs 2 or Re 1, but I think that is a narrative,” he said.
For buyers focused on outcomes, he offered the example of a business spending Rs 20,000 to generate a sales lead and asking whether AI could deliver it for Rs 15,000, with the lead’s cost determining the value of the service.
“So, there are some players who are bold, taking an outcome-based positioning in the market, which I think is the right future,” Bhatia added.
He said checking sales leads and collecting payments are relatively low-risk tasks that sales agents handled earlier, making them suitable for result-based pricing.
Exotel focuses on customer service automation, where Bhatia sees long-term value in handling routine requests such as checking an order or account balance, while ensuring a human can intervene when a customer becomes frustrated.
He said insurance and institutional research salespeople need different skills and are paid differently, arguing that voice AI pricing should also reflect the work it does.
“So, how can it be that everything can be priced to a per minute and everybody is chasing that to the bottom?” he said.
He, nevertheless, distinguished voice AI applications from the underlying communications infrastructure, which he described as a business built around scale and an important part of Exotel’s operations.
The cost of outcome delivery
Bhatia said buyers also need to budget for monitoring and improving AI agents, beyond the quoted price.
He said Exotel focuses on helping its systems remember past conversations, decide whether a human or AI should handle a query and check the quality of a conversation.
“I think you need continuous improvement, evaluation and a process owner who takes accountability. You need somebody who is orchestrating what the bot should do and manage how it works, and somebody to keep sampling calls and listening to them,” he said.
That ongoing evaluation includes checking conversations and improving the system, while deployment time also affects the overall cost, according to Bhatia.
“People are not budgeting for those things. Also the time it takes, like as I said, people build in their budgets that 60% of the conversation will be automated in three months. Those are wrong assumptions,” he said.
Bhatia said automating conversations around one specific customer request could be achievable in that period, suggesting that businesses begin with a simple task and expand gradually as they test its performance.
He suggested moving from 2% automation to 5% and then 10%, giving businesses room for evaluation before taking on more complicated customer requests.
Exotel has already moved parts of its business from charging per user towards consumption measured by minutes or conversations, while also looking at how to link payments to business outcomes, according to Bhatia.
“We are also positioning outcomes in certain cases with some of our design partners that I only want to charge for outcome...I think future is outcome-based,” he said.