The enterprise adoption of AI, both in India and globally, has shifted away from employee-led initiatives to leadership-driven programmes, with senior executives and middle managers now steering top-down deployment of the technology.
For instance, over 80 per cent of Indian business leaders said that they are already familiar with AI agents in comparison to 66 per cent of employees, according to country-specific findings reported by Microsoft India in its annual Work Trend Index (WTI) 2025 report last month.
Globally, 79 per cent of leaders believe that AI will accelerate their careers as opposed to 67 per cent of employees. It is in stark contrast to last year’s WTI report by Microsoft which showed that employees were the key drivers of AI adoption.
This shift in momentum is what is propelling generative AI pilots into production, according to Himani Agrawal, Chief Operating Officer, Microsoft India and South Asia. “When it starts with the leaders, it is serious adoption, it is serious ROI and things which mean business returns. That’s how it gets perceived, conceptualised, and adopted,” Agrawal told The Indian Express on the sidelines of the Microsoft WTI event in Noida recently.
Her remarks come against the backdrop of an ongoing debate over who is responsible, and who should be responsible, for the adoption of AI tools within an organisation. Last month, Coinbase CEO Brian Armstrong said that engineers who refused to sign up to use AI coding tools were immediately fired. The crypto exchange’s mandate drew criticism for disregarding individual choice.
The Indian Express sat down with Agrawal as well as Manpreet Singh Ahuja, Chief Digital Officer, PwC India; and Rajesh Kumar R, Executive Vice President and Chief Information Officer, LTIMindtree, to discuss leadership-driven AI transformation, the flaws in generative AI pilots, AI fluency as a core skill, and more.
Leadership-led AI adoption
When asked why this approach was better than employee-led initiatives, Singh pointed to a market trend from 2023-24 where most people were skeptical of AI. “The conversation was, ‘please come do what you can do, as long as you deliver the ROI back within the six-month or one-year period’. The consultant was sceptical and the customer was sceptical,” he said.
Story continues below this ad
“From there, the mindset shifted because the leadership is getting more engaged, they are using it at a personal level. The ecosystem is beginning to use it. They are a lot more sure and confident,” Singh added.
A recent survey of 3,000 professionals by hiring platform Indeed found that middle managers are leading AI upskilling with 49 per cent of the respondents aged between 35 and 54 stating that they are actively seeking more AI training, compared to 41 per cent of respondents aged between 18 and 24.
Meanwhile, Kumar said that IT services company LTIMindtree wants to be “seen as an AI company solving our customers’ problems with AI, and we need to be adopters internally.”
Scaling generative AI pilots
A recent study conducted by Massachusetts Institute of Technology (MIT) grabbed headlines after reporting that 95 per cent of 300 US-based firms that had invested somewhere between $35 billion to $40 billion in generative AI, saw little to no returns largely due to flawed enterprise integration of AI.
Story continues below this ad
However, Singh struck a positive note, comparing AI adoption to venture capital investing: “They put money across 20 startups, 14 of them may not work out but the six that do will create disproportionate value. Now, if you apply the same mindset to a company looking to adopt AI, where you let those 20 ideas come to life and provide seed capital, you will have the five superheroes come down to the table and unlock value which is beginning to happen.”
When asked about scaling AI pilots within LTIMindtree, Kumar said that the company has fostered a culture of experimentation which encourages the concept of failing fast.
“You fail quick enough so that you don’t spend too much money. If you start counting every failure, it might add up to a percentage but overall it will be a successful journey,” he said, citing the example of LTIMindtree’s project to give every employee a digital companion. Kumar also said that too many constraints can derail AI adoption and limit innovation.
We allocated sufficient resources to create a buffer zone for employees to experiment freely with AI and build solutions without worrying about consumption limits, he further said. “When we get to a level of maturity, then the optimisation comes in, then our wide rollout comes in. That’s the method,” he added.
Story continues below this ad
Data readiness as a bottleneck
Successfully scaling AI adoption also requires companies to organise datasets, as per PwC’s Manpreet Singh.
“When you try to scale, you want the underlying datasets to be strong for you to be able to get foolproof responses when you’re dealing with the customer or employee. While AI agents can function as retrieval tools regardless of whether you have the data or not, if you want to hyperpersonalise that conversation back to every customer, you need customer data to know what they like and don’t like,” he said.
“As you solve that data layer, I think AI will become more and more real,” Singh added.
AI agents and ROI
Over 59 per cent of Indian business leaders are already deploying AI agents to automate workflows across entire teams, according to Microsoft’s WTI 2025 report, with 93 per cent of leaders expected to deploy AI agents to extend workforce capabilities in the next 12-18 months.
Story continues below this ad
Widely touted as the next AI frontier, agentic AI solutions are also compute-intensive and expensive to run. When asked how the push for AI agents is impacting Microsoft’s bottomline, Agrawal said, “Any new technology, product, or service comes with high costs when you introduce it on the market. But as you scale adoption, the cost always keeps coming down.”
Prices per token have fallen steadily since the launch of ChatGPT and Azure OpenAI due to the consumption curve and ongoing technology innovation, she added.