Enterprise AI needs measurable gains as adoption grows, say industry leaders

Companies are moving beyond AI pilots and giving employees access to the technology, but the next test is whether that wider use translates into measurable business gains, leaders said at the Bloomberg New Economy Forum in New Delhi.

That will require companies to identify specific business problems, measure the impact of AI and rethink how they operate, they said.

“The next phase really for AI is going to be not experiments, but actually driving outcomes,” said Pragya Misra, head of public policy and strategy for India at OpenAI, during a panel discussion on 'From Pilot to Payoff: Making AI Actually Work'.


Misra cited online car marketplace Cars24, which she said uses OpenAI models for close to a million conversation minutes every month. Nearly 12% of previously lost sellers have returned through an AI-powered customer agent, she said.

Japanese multinational electronics company TDK Corporation is also seeing returns from AI in manufacturing, including efficiency improvements driven by teams on the ground, said Noboru Saito, its president and CEO during the same panel discussion.

More than 40% of TDK’s team members are using AI and the company is deploying more than 2,000 AI agents, Saito said. It also set up a dedicated global AI organisation called TDK SensEI more than two years ago.

For companies investing in the technology, adoption alone cannot establish its value, said Mohamed Kande, global chairman of PwC.

“It’s not just the use; it’s about the outcome,” he said.

Also Read: Anthropic warns small AI models could self-replicate across smart devices by year-end

Rethinking how companies operate

Kande said companies were beginning to look beyond employee productivity to examine how AI could change the way they run their businesses.

Much of enterprise AI adoption remained superficial, with companies building individual agents for specific tasks instead of rethinking their operations, said Kai-Fu Lee, CEO of Chinese AI startup 01.AI and chairman of Sinovation Ventures.

He argued that AI transformation should start with the CEO rather than the chief information officer, because senior leadership was better placed to identify where the technology could have the biggest impact.

Lee said 01.AI was developing an AI agent for CEOs that could analyse sales reports, records from customer relationship management systems, corporate emails and meeting transcripts. The aim was to help track business performance, identify risks and make decisions.

He also suggested that AI could reduce layers of middle management, shifting responsibility towards employees who directly manage AI agents and deliver results.

Counting the costs and keeping control

PwC had given all its employees access to AI tools about 18 months ago, an investment Kande said was costly. The company wanted employees to understand and use the technology instead of fearing that it would take their jobs.

As usage grows, businesses need to prioritise spending while leaving room for experimentation, he said.

“You don’t get it right the first time. But if you don’t try, you’re never going to get there,” Kande said.

“The journey that many of the companies are going through today is to understand that there is no AI without data, and there is also no AI without governance,” he added.

Companies were investing in data infrastructure, governance, people and security to support adoption, Kande said.

Saito said companies needed both top-down direction and bottom-up experimentation, while balancing the push to adopt AI with governance, risk management and information security.

Data sovereignty also influences how companies deploy the technology. Lee said some customers preferred on-premises AI deployments, where models and data remain within their own environments, because they did not want sensitive information leaving their systems.

Such deployments create implementation challenges, but also business opportunities for companies offering them, he said.

AI race continues despite safety concerns

Despite concerns about costs, security and returns, companies were reluctant to slow AI investments for fear of falling behind competitors, Kande said.

“Despite all the conversation around AI safety, there is no slowdown that we’re seeing in usage of AI and allocation of capital,” he said.

Misra said AI safety remained a priority for OpenAI, adding that the company had slowed certain activities and paused reinforcement learning training for a period.

“If at any point we find that the frontier is getting so far ahead that we’re not being able to deploy these models in a safe way, we will certainly slow down and recalibrate,” she said.