Infosys's Anant Adya on why enterprise AI pilots fail to scale

Enterprise spending on artificial intelligence shows no sign of easing, but new research from Infosys suggests most large companies still cannot show what it is returning. In The AI ROI Gap: Turning Ambition Into Enterprise Value, published on 25 August, 72 per cent of the senior executives surveyed said fewer than a quarter of their AI pilots had scaled to enterprise-wide production and delivered their intended return.

Anant Adya, Executive Vice President and Head of Americas Delivery at Infosys, answered written questions from The Datatech Times on what separates the organisations turning AI into measurable gains from those that are not.

The research found that most organisations are not short of conviction. Some 87 per cent of respondents agreed that AI investment is a strategic necessity, and 80 per cent see it as a critical driver of new revenue. Yet a quarter said the return on their AI investments was below expectations, and another 15 per cent said it was too soon to tell.

"The biggest differentiator is not technology sophistication but execution discipline," Adya said. "The research shows that organisations achieving meaningful AI returns have clearer success metrics, stronger alignment between business goals and use cases, and more structured approaches to measurement and scaling."

He pointed to the scaling figure as the most revealing in the report. "The primary reason is not technical performance but the absence of a clear business case and measurable outcomes from the start," he said. "Successful organisations begin with specific problems to solve, define ownership, and establish metrics before deployment."

Measurement is the other weak point. Two-thirds of respondents said their organisation struggles to measure the return generated by AI, and 46 per cent lack a centralised KPI framework for it. Only a quarter formally track speed to market.

"While many companies focus on revenue, AI's strongest demonstrated benefits today are speed to market, operational efficiency, and cost savings," Adya said. "Organisations that track these gains are better positioned to demonstrate value."

A workflow accelerator

The study's modelled impact scores, which control for industry, company revenue and the scale of AI use, put speed to market and cost savings at the top and revenue gains last. Even among firms that have fully scaled AI across their key operations, revenue impact reached only about half the level of speed to market.

"These findings indicate that AI is primarily functioning as a workflow accelerator," Adya said. "Organisations are using it to reduce manual tasks, automate routine processes, speed content creation and analysis, improve access to knowledge, and help employees complete work more efficiently. In many cases, AI is enhancing existing workflows rather than creating entirely new revenue streams."

Operational teams appear to benefit most, he said, because the metrics they already track, such as cost savings, employee productivity and efficiency, tend to show measurable results quickly. "The broader takeaway is that AI creates value by removing friction from everyday work."

People and process

The report also records a rising workforce risk alongside usage. Some 65 per cent of respondents estimated that at least one-fifth of employees engage in AI-generated "workslop", and 36 per cent worry that staff are using unapproved public AI tools that create security risks.

"As AI adoption matures, competitive advantage increasingly depends on how organisations manage people and processes rather than which models they use," Adya said. "The research notes that weak technology is rarely the main reason AI initiatives fail. More often, success is determined by governance, accountability, alignment, and workforce behaviour."

"AI fluency is becoming a critical skill," he added. "Employees must be able to use AI effectively while also validating outputs, exercising judgement, and applying results appropriately."

Cybersecurity and data privacy rank among the top barriers to enterprise-wide returns, and Adya said organisations are responding with safeguards such as human review, approval workflows and transparency requirements.

For enterprises trying to close the gap, his advice was to shift the emphasis. "The most important lesson is to focus less on AI adoption alone and more on execution discipline," he said. "Simply deploying more technology is unlikely to close the gap."

He also argued that leaders should adjust where they expect value to appear first, and treat workforce readiness as a strategic priority. "Training, oversight, accountability, and responsible-use policies are essential," he said. "Organisations that combine strong governance with employee enablement are more likely to scale AI successfully and create lasting competitive advantage."

The report recommends that companies ground AI business cases in the areas already producing measurable gains, track speed to market as a formal KPI alongside revenue and cost, and pair clearer business cases with regular training on safe and effective AI use. It follows the Infosys Knowledge Institute's Future of Work research, released in July, which found that 74 per cent of US respondents save more than three hours a week using AI tools, against 67 per cent globally.