 "artificial intelligence, AI")
By most visible measures, 2025 was a breakout year for artificial intelligence (AI). AI tools became mainstream among individual users, developers, and creators, reshaping how people write, code, design, and search. Yet, inside large enterprises, the story is far less definite.
Despite widespread experimentation and investment, AI has struggled to translate early adoption into meaningful, scalable business value, exposing a widening gap between enthusiasm and impact.
The McKinsey Global Survey on the state of AI shows that nearly 90 per cent of enterprises globally are using AI in some part of their organisation. However, when asked how many have fully scaled those use cases, the number drops to 7 per cent.
Findings from the MIT Project NANDA report, “The GenAI Divide: State of AI in Business 2025”, reinforce this disconnect. Despite $30-40 billion in enterprise investment in generative AI, nearly 95 per cent of organisations are seeing zero return. The outcomes are starkly divided across both buyers — enterprises, mid-market firms, and small and medium businesses — and builders, including startups, vendors, and consultancies. Just 5 per cent of integrated AI pilots are extracting millions of dollars in value, while the vast majority remain stuck, with no measurable impact on the profit and loss statement.
Noshir Kaka, senior partner at McKinsey & Company, agrees with these findings. “The gap between AI experimentation and real value comes down to three core issues,” he says. First, most enterprises fail to reimagine their business domains. AI is treated as a tool for incremental efficiency or cost reduction, rather than a lever for step-change innovation and growth. “The companies that succeed aim for 2x–10x transformation, and this rethinking must be led by domain leaders, not technology teams.”
Second, he says, organisations start with technology modernisation instead of workflows. New tools are adopted without redesigning processes end to end, leading to long implementations and limited impact. Tech stacks need to be modular, scalable, secure, and built with the right total cost of ownership, but only after the business problem is clearly redefined, he says.
“Finally, enterprises underestimate organisational rewiring. Scaling AI requires new governance, faster decision-making, sustained upskilling, and leadership ownership,” he explains, adding that if leaders cannot use and demonstrate AI themselves, adoption stalls. “Together, these three pillars form a multiplicative equation. Address all three, and AI can finally scale.”
With return from AI still elusive, enterprises are already reshaping cost structures, most visibly through workforce adjustments. Over the past year, several large technology and consulting firms have announced layoffs, often citing efficiency gains from automation, AI-assisted coding, and back-office optimisation. While companies stop short of attributing job cuts directly to AI, the subtext is clear: In the absence of revenue upside, they are using AI first as a cost lever rather than a growth engine.
According to Layoff.fyi, which tracks layoffs, primarily in the tech industry, around 122,549 tech employees were laid off in 2025 from about 257 companies globally.
AI@2026
With the return on investments (RoI) yet to be counted as significant, many believe 2026 could be a defining year for enterprise AI. A survey on the state of enterprise AI adoption by ISG, a global AI-centred technology research and IT advisory firm, found that the number of use cases in production have doubled since 2024.
According to the report, “Despite heavy investment, many enterprises report an AI value gap: Efficiency gains materialise, but business growth impacts lag. In ISG surveys, nearly half of enterprises expect meaningful AI-driven growth only in 2026 or beyond.” The report added that “to accelerate adoption, leaders should focus not only on cost savings and automation but also on growth-oriented use cases – new product design, customer experience and industry-specific reinvention.”
Mark Roberts, head of Capgemini’s AI Futures Lab, points to a shift in enterprise behaviour. Companies, he says, are moving away from running multiple experiments, and are increasingly focused on making AI work at scale. “We are already seeing conversations with customers around integrating these proofs of concept into core systems,” he says. “On returns from AI, many agree it has not delivered what it is capable of, but we expect to see tangible returns in 2026.”
As organisations move towards scaling, data integration is turning out to be a critical bottleneck.
“A trend that has emerged in the last 18 months is that AI in the enterprise only works — and makes sense — when it is grounded in enterprise data,” says Christian Kleinerman, executive vice-president of product at Snowflake, a cloud-based data platform. “If you do not connect your data to AI, all you have is ChatGPT, which will not know how to answer questions about your products or customer services. It is enterprise data that gives AI its context.”
India's AI story so far
As enterprises grapple with these issues, one overarching theme for 2026 appears to be “sovereign AI”. Amid geopolitical uncertainty, tech sovereignty is fast becoming a key aspect, with the need to be resilient. This is evident from the investments worth billions of dollars that major tech firms have announced for India to cater to the demand for sovereign AI. Amazon, Microsoft, and Google have committed investments totalling around $70 billion in India.
It is in this backdrop that India is set to host the Global AI Summit in 2026. The summit is not merely a diplomatic milestone, but a signal of intent: That India aims to shape the next phase of AI adoption around public infrastructure, responsible deployment, and domestic capability-building.
In the run-up to the summit, more than 200 events have been organised across the country. Over 200,000 students and researchers have participated in these events so far.
Under the IndiaAI Mission, the flagship scheme through which companies have brought graphics processing units (GPUs) to India and begun developing indigenous large language models, 2025 marked a critical shift. Under the Rs 10,372 crore mission, the first multi-modal LLM, BharatGen, was announced, as were four additional startups to build LLMs under the mission.
Meanwhile, the number of GPUs onboarded and brought into the country under the IndiaAI Mission has crossed 38,000. That’s nearly four times the initial target of 10,000. Similarly, the number of datasets on the AIKosh platform – a source for datasets, models, and toolkits to enable AI innovation – crossed 3,000 datasets, and the number of AI models across 20 sectors touched 243.
With 2025 having laid the foundation for AI adoption, 2026 could well be the year when a discernible structure begins to emerge.
