The trust dividend: Why India’s most disciplined AI teams are pulling away from everyone else

Ninety-one percent of organisations surveyed in India say they’re increasing AI spending over the next year. That number, on its own, tells a familiar story of a market racing to adopt. What the underlying report actually documents is stranger, and more consequential: spending more on AI is not what separates the companies getting a return on it from the companies that aren’t.

What separates them is trust — specifically, whether an organization has built the governance, data quality and auditability infrastructure to make its AI systems explainable, correctable and repeatable.

The second annual Data and AI Impact Report, published by SAS with research from IDC, surveyed 2,699 decision-makers across 28 countries and four industries — banking, insurance, life sciences and the public sector — about how their organizations manage, govern and profit from AI. Its headline finding is blunt: organisations with strong trustworthy-AI practices were 15 times more likely to report strong or high ROI from their AI projects than organizations without them.

That 15x multiple isn’t a rounding effect of scale or budget. Organisations with the strongest governance and data-quality practices reported at least double the ROI of their peers on AI deployments generally, and fewer than one in twenty laggard organisations could say the same about strong returns. The report frames this as a widening divide rather than a static gap: 85% of the AI leaders identified in the survey are increasing their investment in trustworthy AI practices by more than 10% this year, which means the distance between the disciplined and the undisciplined is set to grow, not close.

The autonomy problem
The report’s most uncomfortable finding is about what happens once AI stops just recommending things and starts acting on its own. Trust in generative AI systems among respondents sits at 76%. Trust in agentic AI — systems that take multi-step actions with less human sign-off — falls to 66%. That ten-point drop is the report’s clearest evidence that autonomy and confidence move in opposite directions once systems are given more room to operate independently.

That erosion of confidence shows up downstream as a habit: overriding the machine. Across the surveyed organizations, 97.2% of users override AI-generated recommendations in at least some cases — not a fringe behavior but close to universal. The report’s explanation for why is telling. The single biggest reason employees overrode an AI recommendation, regardless of whether the output was actually correct, was that the system could not explain how it reached its decision. Not accuracy. Not speed. Explainability.

“As AI becomes more autonomous, organisations face a new challenge: maintaining confidence in systems people don’t fully understand. Our findings show that stronger oversight, explainability, accountability and data foundations are becoming prerequisites for scaling AI successfully,” said Chris Marshall, Vice President, IDC.

In India specifically, the two most common reasons employees gave for overriding AI outputs were an insufficient explanation behind the recommendation and a lack of adequate business context — echoing the global pattern almost exactly. The report frames this override behavior as more than a nuisance: every manual correction is time, productivity and confidence that the organization spent building an AI system in order to lose back again.

The infrastructure nobody wants to fund
Underneath the trust problem sits a data problem, and the report’s numbers here are the starkest in the entire study. Only 17.5% of enterprises have a fully optimized data infrastructure mature enough to support the demands of agentic AI. The other roughly five in six are running increasingly autonomous systems on data foundations that were not built for the job.

Organisations with an optimized data foundation are four times more likely to expect strong ROI from AI projects — and six times more likely to mandate the data-quality and explainability controls that build trust in the first place.

In India, the message appears to be landing, at least directionally: 69.4% of surveyed organisations named data quality and governance as one of the most important factors in their AI strategy, and the financial services and technology sectors were singled out as showing the clearest year-on-year progress in maturity and infrastructure. But naming a priority and funding it are different things, and the report’s global infrastructure numbers suggest most organisations — in India and elsewhere — are still investing in AI capability faster than they are investing in the data foundation that capability depends on.

“State-of-the-art agents can have error rates that exceed 25% on complex tasks — which is unacceptable in high-stakes decision-making. In order to achieve accuracy and repeatability, organisations must embed domain expertise into agentic workflows, while keeping people at the center of governance and oversight,” said Bryan Harris, Chief Technology Officer, SAS.

India’s spending curve is steepening
Whatever the governance gaps, the appetite for AI investment in India is not slowing down — if anything, it is concentrating into bigger bets. Among surveyed organisations, 91.8% expect AI spending to increase over the next 12 months, roughly in line with global sentiment. The more telling shift is in the size of those increases: the share of Indian organizations planning to raise AI spending by more than 20% jumped from 5% in 2025 to 27.3% in 2026 — a more than fivefold increase in the population of aggressive spenders in a single year.

“The findings reflect a broader shift among organisations in India toward strengthening the foundations required to support trustworthy and explainable AI. With over 90% of surveyed organizations planning to increase AI investments, establishing trust in AI systems will be critical to realizing lasting business value,” said Noshin Kagalwalla, VP – Public Sector, APAC & Managing Director, SAS India.

Read against the ROI findings, that spending curve cuts two ways. It could mean Indian organisations are getting more confident and putting real money behind AI systems that are starting to work. Or it could mean a growing share of India’s AI budget is about to be spent by organisations that haven’t yet built the governance layer the report says is the actual precondition for a return — in which case the spending increase mostly buys more overrides, more distrust and more of the 25%-plus error rates Harris describes on complex tasks.

How the divide plays out by industry
The report’s four focus industries show trustworthy AI maturity translating into different competitive dynamics depending on the sector’s regulatory exposure and risk tolerance.

Strip away the framing and the report is making a fairly specific argument: the technology layer of AI has stopped being the differentiator. Everyone surveyed has access to broadly similar models and tooling. What separates the 62%-ROI organisations from the 4%-ROI organisations is not what they bought — it’s whether someone built the unglamorous scaffolding around it: data pipelines clean enough to trust, decision logs detailed enough to audit, and enough domain expertise embedded in the workflow that a human reviewer can tell, at a glance, why the system did what it did.

For India, where investment intent is running well ahead of the 17.5% of enterprises with data infrastructure actually ready for agentic AI, that argument lands as a warning as much as a validation. The organizations naming data quality and governance as a top priority — 69.4% of those surveyed — appear to understand the diagnosis. Whether the 27.3% of Indian organizations about to significantly increase AI spending direct that money toward the cure, or toward more of the same underlying infrastructure, is likely to decide who is telling the ROI story a year from now, and who is explaining the override rate.

Findings are drawn from SAS and IDC’s second annual Data and AI Impact Report: The New Economics of Trust, based on a global survey of 2,699 decision-makers with knowledge of, or influence over, their organization’s data and AI initiatives, conducted across 28 countries and four focus industries — banking, insurance, life sciences and the public sector.