Speaking to CNBC-TV18, Sylvain Duranton, Managing Director & Senior Partner at BCG
“The peak is ahead of us,” Duranton said, adding that he sees no signs of companies slowing their AI investments despite growing scrutiny over returns.
His comments are backed by BCG’s latest survey of 1,800 C-suite executives conducted ahead of the World Economic Forum in Davos. The survey found companies expect to double the resources dedicated to AI transformation between 2025 and 2026, with AI spending averaging around 2% of revenues.
According to the survey, 94% of CEOs said they would continue investing in AI even if they failed to achieve the expected returns in 2026, while 27% said they would increase spending because it would suggest they had not been ambitious enough.
“I don’t see any signs of hesitation,” Duranton said, adding that many chief executives increasingly view AI transformation as part of their long-term leadership legacy.
As AI adoption accelerates, however, companies are becoming far more conscious of the economics of using large language models.
“One of the key concerns now is the cost of tokens, which is becoming very real for many companies,” Duranton said.
Rather than relying exclusively on the largest commercial AI models, enterprises are increasingly evaluating open-weight and smaller models for routine tasks while reserving frontier models for more complex workloads.
“Today, we are in a situation where everyone is using a fighter jet to go and fetch their kids from school,” Duranton said, arguing that many enterprise applications do not require the most powerful and expensive AI models.
The shift reflects a broader change in enterprise AI strategy as organizations seek the right balance between performance, cost and security. Depending on their operating markets, some companies continue to favor Western AI models, while others are experimenting with Chinese and open-weight alternatives.
Duranton said the conversation has also expanded beyond model performance to include issues such as safety, traceability and cybersecurity.
India leads AI adoption, but monetizing it remains a challenge
While enterprises are rapidly embracing AI, converting that adoption into measurable financial returns remains the industry’s biggest challenge, according to Nipun Kalra, Managing Director & Senior Partner at BCG X and Global Leader of AI-Financial Institutions Practice.
Kalra said India leads global workplace AI adoption, with around 95% of frontline employees using AI several times a week for work-related tasks.
“The good news is that people are getting comfortable with AI. That’s super positive. But translating that into a P&L gain for an enterprise is where the gap is,” he said.
BCG’s research also shows that only 14% of organizations globally measure AI initiatives against clear profit-and-loss metrics, making it difficult to assess whether investments are delivering business value.
The leadership gap is another challenge. While 72% of CEOs globally personally lead AI initiatives, the figure falls to 55% in India, suggesting many AI programs continue to be driven primarily by technology teams rather than business leaders.
Kalra argued that organizations treating AI as a business transformation programme, rather than a technology deployment, are more likely to generate meaningful returns.
From AI pilots to enterprise-wide transformation
Kalra said many companies are diluting the impact of AI by launching numerous pilot projects instead of fundamentally redesigning business processes.
“The organizations most likely to succeed are the ones that focus on only a few priorities and completely rewire an entire function, customer journey, or business process end-to-end,” he said.
He also stressed that enterprises must build AI capabilities internally instead of relying indefinitely on consultants.
“You may need an external catalyst to get started and help build that capability. But if you don’t have that internal muscle, it’s very difficult to keep pivoting, iterating and improving your core operations every day, every month and every year. That internal capability is non-negotiable.”
Duranton added that AI is already reshaping software engineering itself, with developers increasingly building systems that generate, test and validate code rather than writing every line manually.
“This is not incremental change,” he said.
For enterprises, the next phase of AI adoption is likely to be driven less by experimentation and access to the largest models than by choosing the most cost-effective model for each workload, building internal capabilities and demonstrating measurable business outcomes.
