Artificial intelligence moves from experimentation to mainstream business adoption, companies are facing a more demanding question: are their AI investments delivering measurable returns? For business leaders, the challenge is increasingly less about whether to adopt AI and more about translating that investment into tangible improvements in productivity, efficiency, decision-making and financial performance.
In this exclusive TechBullion interview, Nithin Mummaneni, CEO of Infinity Loop, discusses the growing pressure on organisations to demonstrate measurable AI ROI and explains why promising AI initiatives can struggle to progress beyond the pilot stage. He explores how businesses can evaluate AI performance, identify meaningful success metrics, build stronger investment cases and integrate AI into existing workflows.
Mummaneni also shares insights from Infinity Loop’s work with AI-powered contract intelligence and negotiation, examining how organisations can measure the value created by AI in procurement and commercial decision-making. Looking ahead, he considers the future of enterprise AI, the role of specialised AI solutions and what executives should prioritise as AI becomes increasingly embedded in business operations.
1) Please introduce yourself to our readers. What inspired you to found Infinity Loop, and what gap in the enterprise AI market did you initially set out to address?
I’m Nithin Mummaneni, founder and CEO of Infinity Loop. Before starting the company, I worked as a management consultant across technology, healthcare, aerospace, defense, and consumer products.
Across those industries, I kept seeing the same problem: companies were spending enormous amounts with third-party suppliers without a reliable way to answer a very basic question: Did we actually get a good deal?
There were plenty of tools managing procurement workflows, approvals, and contracts, but very few focused on improving the commercial outcome itself. I founded Infinity Loop to close that gap by giving procurement and business teams the intelligence they need to negotiate better deals and generate measurable value from supplier relationships.
2) How has the conversation around enterprise AI changed as companies move from experimentation towards measurable business results?
The conversation has changed dramatically. Two years ago, there was enormous pressure from boards and C-suites simply to have an AI strategy. Now the question is: What did that investment actually produce?
AI adoption isn’t a business outcome. Enterprises increasingly want to know whether an investment reduced costs, increased revenue or materially improved productivity.
I think we’re entering a much more disciplined phase of enterprise AI. The solutions that get funded and scaled will increasingly be the ones that can connect their technology directly to measurable business results.
3) Why do some organisations struggle to turn promising AI pilots into solutions that deliver meaningful commercial value?
Usually because the problem starts before the pilot.
Companies sometimes select the technology first and define the business problem second. That makes it very difficult to prove value later.
The era of experimenting with AI simply because it’s AI is ending. Before launching a pilot, leaders should be able to answer three questions: What problem are we solving? What outcome do we expect? How will we measure it?
If those aren’t clear at the beginning, proving commercial value at the end becomes extremely difficult.
4) What should business leaders consider when deciding whether an AI initiative is genuinely creating value or simply demonstrating technical capability?
Start with the outcome, not the technology.
Will this increase revenue, reduce costs, or materially improve the performance of a team? And can you measure the difference?
A technically impressive solution isn’t necessarily a valuable enterprise solution. Establish the baseline before implementation, define the outcome you expect to change, and measure against it. That’s how you separate an interesting AI demonstration from a business investment worth scaling.
5) Which metrics provide the clearest indication that an organisation’s investment in AI is generating a meaningful return?
Ultimately, most measurable AI value falls into three categories: revenue growth, cost reduction, and productivity improvement.
The exact KPI depends on the use case. What’s more important is establishing the measurement framework before implementation. If you don’t know the baseline, it’s very difficult to credibly claim ROI afterward.
6) How quickly should companies expect to see measurable returns from AI, and when should leaders reconsider an underperforming investment?
It depends on the use case, but enterprises shouldn’t have to wait indefinitely to understand whether they’re moving in the right direction.
I’d want to see leading indicators within the first few weeks. With Infinity Loop, for example, customers can identify potential savings opportunities within days of uploading contracts. As those negotiations occur, those opportunities can translate into realized hard-dollar savings.
Not every AI solution will generate value that quickly, but every solution should have a defined path and timeline to proving the outcome it was purchased to deliver.
7) What are the most common mistakes companies make when developing a business case for enterprise AI?
The biggest mistake is building the business case around the technology rather than a strategic priority.
You can invent almost any KPI for an AI initiative. But if it isn’t connected to something the organization genuinely cares about, such as revenue, cost, productivity, or risk, it’s going to be difficult to build sustained support.
A strong business case should make the connection between strategic priority, use case, measurable KPI, and expected outcome very clear.
8) Should every AI investment produce a direct financial return, or can strategic and operational benefits provide an equally compelling measure of success?
Not every AI investment needs an immediate hard-dollar return.
Enterprises will ultimately have portfolios of AI capabilities. Some will directly reduce costs or increase revenue. Others will improve productivity, reduce risk, or create strategic capabilities that matter over a longer period.
What matters is accountability. If the return isn’t financial, leaders should still be able to define what strategic or operational outcome they’re buying and how they’ll determine whether it was achieved.
9) How important are data quality and accessibility in determining whether an organisation can generate meaningful returns from AI?
Data matters, but enterprises shouldn’t assume everything needs to be perfectly structured before AI can create value.
The more important question is whether the solution can work with the data the enterprise actually has.
At Infinity Loop, for example, much of the information we work with starts as unstructured contract data. Our technology structures and analyzes that information before applying specialized data, frameworks, and reasoning to it.
AI should adapt to enterprise reality rather than requiring an enterprise to completely reorganize itself before it can see value.
10) What questions should boards and senior executives ask before committing significant capital to new AI initiatives?
I’d ask two questions.
First: Does this support one of our strategic priorities over the next one, three or five years?
Second: What financial, operational or strategic return should we expect, and what evidence suggests this solution can deliver it?
AI shouldn’t sit on an island separate from corporate strategy. It should be held to the same standard as any other significant business investment.
11) What distinguishes companies that successfully scale AI from those that remain caught in repeated pilots and proofs of concept?
Companies that scale AI aren’t just proving that the technology works. They’re proving that it creates enough value to justify deploying it more broadly.
That’s an important distinction.
A successful pilot should answer a commercial question, not merely a technical one. If you can’t demonstrate financial, strategic, or operational value sufficient to support a buying decision, it’s very easy to get trapped in an endless cycle of proofs of concept.
12) How important is redesigning existing workflows when businesses want AI to deliver measurable improvements in productivity and performance?
I’d start with the desired outcome rather than the existing workflow.
AI is inherently disruptive. We shouldn’t assume a process designed around yesterday’s technology needs to survive unchanged.
If AI allows someone to achieve a better outcome in a fraction of the time, redesign the process around that capability. Work backward from the result you want rather than forcing AI into a workflow simply because that’s how the organization has always operated.
13) How does Infinity Loop approach building AI solutions around measurable business outcomes rather than technology adoption alone?
Measurable outcomes are fundamental to how we’ve built Infinity Loop.
We’re not asking customers to judge success based primarily on logins, prompts, or adoption. Our platform is designed to help companies identify and realize material savings through better supplier negotiations.
That gives us a very direct measurement framework: What financial opportunities did we identify? What savings were realized? What costs were avoided?
We want the connection between the technology investment and the business outcome to be as direct as possible.
14) How can companies accurately measure the value created when AI improves contract analysis, procurement or negotiation decisions?
Contract analysis itself is rapidly becoming commoditized. A general-purpose LLM can already summarize a contract reasonably well.
The more valuable question is: What should you actually do with that information?
In procurement, that means identifying the financial opportunity, understanding your leverage, developing the right negotiation strategy, evaluating the deal structure, and accounting for risk.
That’s where specialized data and reasoning become much more important. The goal isn’t to produce a better contract summary. It’s to help someone make a better commercial decision.
15) What lessons can businesses learn from procurement when assessing the financial and operational impact of AI?
Good procurement teams don’t enter major negotiations without a game plan. Companies should treat AI investments the same way.
Before buying, define what you’re trying to accomplish, how it connects to the broader strategy, and exactly how you’ll measure success afterward.
In other words, build the playbook before making the investment, not after you’re trying to justify it.
16) How should enterprise buyers assess AI vendors’ ROI claims and distinguish credible projections from unrealistic promises?
Don’t accept the vendor’s ROI framework. Establish your own.
Tell the vendor exactly what outcomes you’ll use to evaluate the investment and ask them to show what you should reasonably expect against those metrics and what evidence supports that expectation.
That turns an ROI claim in a sales deck into a shared success plan. If a vendor can’t explain what success looks like, how it will be measured, and why they believe they can deliver it, buyers should be cautious.
17) How can businesses balance AI-driven automation with human judgement when making complex or commercially significant decisions?
For complex, high-value decisions, I believe strongly in keeping humans in the loop.
Most companies aren’t ready to hand a multimillion-dollar strategic supplier negotiation entirely to an autonomous agent, and I’m not convinced removing the human should be the objective.
The better model is to give people dramatically better intelligence: the right data, insights and negotiation playbooks at the right moment. AI can improve the speed and quality of decision-making while the human retains judgment and accountability for the final commercial decision.
18) Will specialised enterprise AI solutions ultimately deliver greater business value than general-purpose AI tools?
In many cases, yes, particularly where the problem requires specialized data, domain expertise, and a measurable outcome.
General-purpose AI is extraordinarily powerful, but it’s designed to do many things. Specialized solutions can combine that underlying capability with proprietary data, workflows, reasoning, and measurement frameworks built around a specific business problem.
Ultimately, enterprises will use both. But specialized AI companies won’t survive simply because they’re specialized. They’ll survive because they can prove they create more value.
19) How do you expect the way companies measure AI performance and ROI to evolve over the next three to five years?
I expect AI measurement to become much more formalized and financially disciplined.
There will be thousands of AI solutions that look increasingly similar on the surface. That will make technical capability alone a weaker differentiator and put much more pressure on vendors to demonstrate measurable value.
The question will shift from “Does this use AI?” to “Does this materially improve the economics or performance of our business?”
That will increasingly determine which AI vendors get purchased, scaled, and renewed.
20) What advice would you give leaders who have invested heavily in AI but are still struggling to demonstrate measurable returns?
I’d stop buying more technology for a moment.
Go back to the business objectives. What did you actually expect AI to accomplish: increase revenue, reduce costs, improve productivity or reduce risk? Then look at the investments you’ve already made and determine why they haven’t delivered against those outcomes.
Do the postmortem. Establish clearer selection criteria and measurement frameworks. Then concentrate investment on the areas where AI can genuinely move the needle.
Sometimes the problem isn’t that AI failed. It’s that the organization never defined what success was supposed to look like.
