In the Race for AI ROI, Domain Expertise is Becoming Ultimate P&L Metric

For the past two years, enterprises have rolled full steam ahead when it comes to AI experimentation. Companies have launched copilots, deployed chatbots, and tested generative AI across functions. 

But as they now move from pilots to deployment, they want to measure their returns on AI investments. However, tracking hard numbers like revenue lift or margin expansion is notoriously fuzzy given how AI has seeped into every vertical, if not every workflow.

According to a recent study commissioned by Cognizant and Dun & Bradstreet, 80% of enterprises in India are prioritising technology investments for revenue growth, while only 12% have a fully consolidated, enterprise-wide view of technology spending needed to accurately measure returns.

"Up until about 12–18 months back, everything was experimentation," Nitesh Mirchandani, Chief Business Officer of technology services company Mindsprint, tells AIM. "Now, the industry has pivoted from experimenting to productionising AI. It's about measuring ROI—not just on paper, but in reality."

Enterprises are now changing the way AI projects are evaluated. Instead of asking whether a model can generate content or answer questions, businesses are increasingly asking whether it can shorten procurement cycles, optimise inventory, reduce manual effort, or improve operational efficiency.

How Should AI ROI Be Measured? 

For Hinduja Global Solutions (HGS), that shift has become central to how it approaches enterprise AI deployments.

“We call it 90-day ROIX, which is a play on the idea of intelligent experiences, return on intelligent experiences,” HGS Global CEO Venkatesh Korla explains.

HGS embeds a forward deployment engineering team inside client organisations to identify business processes that are ready for transformation. Instead of attempting to automate everything, the company prioritises a handful of high-impact initiatives.

“We would then identify the top five or 10 items… pick the top one or two that have the most value with the fastest return on investment,” he notes.

The emphasis on measurable gains over technological novelty reflects a broader evolution in enterprise AI. That change is particularly visible in industries where outcomes are easier to measure.

Manufacturing, logistics, mining, and supply chain operations have long relied on operational metrics such as utilisation, turnaround time, inventory levels, and scheduling efficiency. Improvements in these areas translate directly into financial gains, making them some of the earliest environments where AI's business value can be clearly demonstrated.

Optimising Operations

Bengaluru-based AI and mathematical optimisation startup Optiflux is working precisely in these operational settings.

The company recently worked with a mining enterprise that had to coordinate thousands of workers across mines, rail networks, and ports while complying with complex workforce regulations. A disruption in scheduling, such as a train driver becoming unavailable, could delay shipments worth tens of millions of dollars.

To address the problem, Optiflux built an optimisation engine capable of dynamically reallocating qualified workers while maintaining compliance requirements. According to the company, the solution improved workforce compliance from 85% to 97.6%, significantly reducing operational disruptions.

"We don't claim to be an AI company. We're a supply chain optimisation company first," Co-founder and COO Vivek Prakash says, arguing that the ROI on AI tools should be measured in how they solve complex operational problems.

Measuring Domain Expertise

The common thread tying AI ROI is that it is measured on business outcomes—such as cost reduction, revenue growth, and risk mitigation—and positioning the technology as merely an underlying tool.

Mindsprint's own deployments reflect that philosophy. The company has been building AI solutions across procurement, finance, and supply chain operations, where the technology automates invoice processing, assists procurement teams in supplier evaluation, and streamlines enterprise workflows. Its Mindverse platform offers plug-and-play tools for outcome-based service operations, while its agentic AI portfolio spans procurement management and cybersecurity risk management.

"The conversation is no longer about another proof of concept," Mirchandani says. "Customers are asking what business outcome they're going to achieve."

The company believes that this shift is also exposing another challenge in enterprise AI adoption: domain expertise.

A recent study conducted by HFS Research found that nearly half of business leaders surveyed said technology providers lacked sufficient industry expertise to solve business-specific problems. The finding suggests that access to large language models is rapidly becoming commoditised. Most enterprise software companies can integrate foundation models from providers such as OpenAI, Anthropic, Google, or open-source alternatives.

What is becoming more difficult to replicate is a deep understanding of industry workflows. Understanding procurement cycles, mining operations, airline crew scheduling, or manufacturing constraints may increasingly matter more in the near future than selecting one AI model over another.

That philosophy is reflected in Optiflux's own approach. Instead of beginning with a generic AI product, the company has spent its early years solving bespoke optimisation problems across industries before deciding which capabilities could eventually become repeatable software products.

It also explains why many enterprises are shifting their conversations away from AI features and towards measurable outcomes.

Operational AI is easier to justify. Better workforce allocation reduces overtime. Smarter procurement lowers purchasing costs. Improved scheduling increases asset utilisation. Better inventory planning reduces working capital.

That growing emphasis on accountability is also beginning to influence commercial models. According to Mindsprint, customers are increasingly exploring engagement models linked to business outcomes and measurable ROI, even though outcome-based pricing is still evolving.

The transition is unlikely to happen overnight. But as AI investments become larger and boards seek greater accountability, enterprises are becoming more selective about where they deploy the technology.