The AI boom may have encouraged companies globally to move quickly from experimentation to deployment, but a string of high-profile rollbacks has highlighted the risks of scaling the technology without adequate data, governance and operational readiness. In India, enterprises have generally moved more cautiously, helping them avoid some of the costly reversals seen overseas, but the focus is now shifting from adopting AI to ensuring that it delivers measurable business value.
Earlier in May, Starbucks discontinued an AI inventory tool deployed across its stores in North America after repeated errors led to mislabelling and inaccuracies in item counts. US automaker Ford also said last month that it had re-hired 300 veteran quality inspectors after AI-powered cameras used to detect quality problems in vehicles failed to meet its standards. McDonald’s, meanwhile, removed AI systems being tested for order processing after the technology repeatedly misinterpreted customer orders and ingredients.
Indian enterprises have largely avoided such large-scale pullbacks, partly because fragmented databases, legacy technology and integration challenges have made companies more cautious about moving AI projects from pilots to production. “In India, the clearer pattern is tighter scrutiny, slower scaling, and selective movement from pilots to production,” Biswajeet Mahapatra, principal analyst at Forrester, said. Enterprises are reassessing AI initiatives where the return on investment, governance, data readiness or integration remains weak, he added, but this does not amount to a wholesale pullback from the technology.
Instead, Indian companies are increasingly moving towards what industry executives describe as AI discipline, with greater emphasis on use cases where the benefits can be measured and accountability for deployment is clearly defined. “The rollback here is not from AI but from unmeasured AI,” Kanakalata Narayanan, vice-president of engineering at Ascendion, said. Boards are increasingly asking where AI will translate into productivity, cost savings, risk reduction, better customer experience or revenue rather than simply whether a pilot can be launched, she said.
This shift is particularly important as enterprises confront a wide gap between AI experimentation and business outcomes. Recent research has indicated that as much as 95% of enterprise AI initiatives are yet to deliver their intended outcomes at scale. Salesforce research also found that while Indian organisations showed a greater willingness to adopt AI than their global counterparts, 38% of Indian respondents said they had experienced an unsuccessful AI pilot, compared with a global average of 28%. At the same time, Indian employees were 45% more likely to say AI was core to their company’s workflows.
The challenge, therefore, is increasingly one of execution rather than access to AI. “In India, we continue to see strong momentum, with enterprises moving beyond experimentation and increasingly embedding AI into core business processes. The challenge is not AI adoption, but AI value realisation,” Amit Kirti, strategy and execution services and AI leader at GDS EY-Parthenon, said. Workforce readiness, leadership alignment, trust, change management, governance, data quality and the willingness to redesign workflows can all determine whether an AI deployment delivers its expected returns.
This is also creating a larger role for IT services firms and technology architects, as companies look for help in cleaning and integrating legacy data, redesigning processes, putting governance frameworks in place and taking successful pilots into production.