The AI ROI problem: Why Indian enterprises cannot yet prove return on their biggest technology investment

By Abhishek Rungta, Founder and CEO, INT (Indus Net Technologies)

India’s enterprise AI ambition is not in question. Nearly half of Indian enterprises, 47%, now have multiple Generative AI use cases live in production, and 76% of Indian business leaders believe GenAI will have a significant business impact, according to the EY-CII AIdea of India: Outlook 2026 report. Budgets are being committed. Vendors are being shortlisted. Projects are being launched across functions and verticals with an urgency that signals genuine organisational conviction.
And yet, beneath that momentum, a structural problem is quietly compounding. Indian enterprises are investing in AI at scale. Most of them cannot prove that it is working.

IBM’s Q4 2025 Think Circle, drawing on discussions with enterprise AI leaders across markets, found that only 29% of executives can measure AI ROI confidently, even while 79% report seeing productivity gains from their AI investments. That gap is the central diagnostic of enterprise AI in 2026. Organisations are experiencing AI. They are not yet measuring it in terms that justify what they are spending.

This is not a failure of ambition. It is a failure of architecture, specifically, the measurement architecture that sits between AI deployment and business value.

Why Productivity Is Not the Same as ROI
The most common way Indian enterprises measure AI success today is operational efficiency: time saved, tasks automated, response speeds improved. These are real gains. They are not ROI.

ROI requires connecting an investment to a financial outcome. It requires a baseline, a measurable change, and an attribution chain clear enough to satisfy a CFO or a board. According to Gartner’s 2026 survey of more than 200 finance chiefs, only 36% of CFOs feel confident in their ability to deliver real enterprise impact from AI initiatives, even as 39% of them list accelerating AI use as a top-five action item for the year. The gap between those two numbers, between AI on the agenda and AI on the P&L, is where the ROI problem lives.

IBM’s APAC AI Outlook 2025, conducted by Ecosystm, found that nearly 60% of surveyed organisations across Asia Pacific anticipate realising the benefits of their AI investments within two to five years, with only 11% expecting returns within the next two years. For Indian enterprises, this has a practical consequence: most current AI investment is being justified on faith in future returns, not evidence of present ones. That is a defensible position in year one of an AI program. It becomes a governance problem in year three.

Three Structural Reasons the ROI Proof Is Missing
The first reason is that AI deployments are being measured against the wrong outcomes. Most enterprise AI projects in India are measured on deployment metrics: adoption rates, query volumes, and reduction in processing time. These are input metrics. They tell you whether the AI is being used. They do not tell you whether the business is better off. A claims processing system that resolves cases 40% faster is impressive. It only creates ROI if that speed translates into fewer outstanding claims, lower indemnity costs, or higher policyholder satisfaction that reduces churn. Without that downstream measurement, the 40% efficiency gain is real but financially unverifiable.

The second reason is that success criteria are defined after the project begins, not before. Research shows that projects with quantified success metrics defined upfront achieve a 54% success rate, compared to 12% for those without clear pre-approval metrics. Most enterprise AI projects in India do not start with a measurable business outcome, a baseline, and a minimum threshold for scaling defined in writing before the first line of code is committed. They start with a use case and an aspiration. The measurement question is deferred to the review meeting, at which point it is too late to answer it rigorously.

The third reason is data fragmentation. Gartner projects that by 2026, 60% of AI projects without AI-ready data will be abandoned, and a survey of 1,203 data management leaders found that 63% either lack or are unsure of having the right data management practices in place for AI. For Indian enterprises, this is compounded by the fact that data accessibility issues are consistently cited as the top barrier to AI success. IBM’s APAC AI Outlook found that 46% of Indian organisations cited data accessibility issues as their primary challenge in realising AI goals, ahead of limited AI skills at 42% and difficulty in integration and scaling at 38%. You cannot measure ROI from AI when the data the AI is acting on is incomplete, ungoverned, or disconnected from the systems where business outcomes are recorded.

What Solving the ROI Problem Actually Requires
The solution is not more AI spending. It is more disciplined AI governance applied before spending begins.

Specifically, it requires three things.

First, every AI initiative needs a business outcome owner, not just a technology owner. The person accountable for measuring AI ROI must be the same person accountable for the business result the AI is intended to create. In most Indian enterprises today, AI projects are owned by technology teams.

Business outcome accountability sits separately, in a different reporting line, with a different performance framework. That structural separation makes ROI attribution almost impossible, because no single person carries accountability for both the investment and the outcome.

Second, baseline measurement must precede deployment. Before any AI project is approved, the enterprise must document the current state of the business metric it is intended to improve. Claims resolution time. Customer acquisition cost. Revenue per salesperson. Support ticket volume. The baseline must be measured, agreed upon, and recorded before the AI system touches the process, or the measurement conversation will be permanently inconclusive.

Third, integration must be treated as the primary engineering challenge, not a post-deployment task. AI that cannot connect to the systems where business outcomes are recorded cannot generate the data trail needed to prove ROI. NASSCOM’s AI Adoption Index found that only 15% of Indian enterprises have aligned their AI strategy with broader corporate strategy, while 67% allocate less than 10% of their IT budget to AI. That combination, low integration investment and strategic misalignment, is precisely why most Indian AI deployments remain technically successful but financially invisible.

The enterprises that will compound their AI investment into a durable competitive advantage are not the ones spending the most on AI. They are the ones that have connected AI to the business metrics that matter, built the measurement infrastructure to track it, and established accountability structures that make ROI a question that can be answered with data rather than faith.

India has the talent, the ambition, and the infrastructure to lead on enterprise AI. The missing piece is not the next model or the next tool. It is the discipline to measure what has already been built, and to build what comes next on a foundation of evidence rather than momentum.