The New Math Behind ROI In AI

The race to deploy AI is heating up, but organisations no longer want just deployment. They’re asking whether copilots, AI agents and automation systems are translating into measurable business value. Both enterprises and startups are under pressure to justify their investments. 

The challenge? Unlike traditional software, AI rarely transforms just one part of the business, making its return on investment (ROI) far more complex to calculate. But the answer also varies from company to company.  

We spoke with AI-native startups, enterprise giants and other scaled-up ventures to understand what ROI in AI actually looks like today. 

Which metrics are being tracked and are factors like productivity and efficiency coming into play in a measurable way rather than remaining subjective? And what happens when businesses find that their AI investments are not delivering meaningful returns? Let’s unpack these questions because in the age of AI, pretty much every company has to grapple with them. 

No One Formula 

Even though there is no universal formula or playbook for calculating ROI from AI, organisations are increasingly measuring AI against the outcomes that matter most to their businesses.

What we learnt is that ROI in AI does not carry the same meaning for every organisation. Across industries, it is increasingly being measured by the business outcomes that matter most — productivity, customer experience, or revenue growth.

For edtech platform upGrad, for example, AI has been embedded across internal knowledge work and customer-facing workflows, including its examination platform and sales funnel. These deployments have generated close to a 5X ROI, largely by reducing manual intervention.

“Using AI has lifted output across our tech, content and marketing teams by 1.4X to 2X, translating into a 7-8X return on our AI spends,” said Anuj Vishwakarma, the CEO of higher education programmes at upGrad.

The company measures AI as output generated per rupee spent instead of cost eliminated, with most productivity gains being used to increase delivery rather than shrink headcount. 

Healthcare companies are measuring success differently. For radiology platform 5C Network, AI’s value lies in improving workflows rather than increasing AI adoption. Faster turnaround time, more consistent reporting quality, better prioritisation of scans and allowing radiologists to spend more time on clinical judgement have become stronger indicators of ROI than tracking employee usage. 

Consumer-facing businesses, meanwhile, are focusing on customer lifecycle metrics.

“Within one to two months, our customers see over 90% AI deflection or resolution rates,” said Dev Ramnane, investor and fractional CBO of SagePilot, a platform that provides AI agents for customer lifecycle management.

According to the company, its AI platform autonomously manages both pre-sales and post-sales conversations with minimal human intervention, reducing support costs by as much as 90% while increasing revenue by 8%-12% without additional advertising spend. 

The BFSI sector is arriving at a similar conclusion, although its metrics revolve around trust and business outcomes.

“Our AI-assisted systems resolve 30%-40% of merchant queries at the first level, while our AI-powered risk models have helped reduce fraud and chargeback losses by nearly 80%. The biggest shift has been moving from measuring AI activity to measuring AI outcomes,” said Manas Mishra, CPO at PayU and Wibmo.

It reflects a broader shift across financial services, where AI is increasingly being measured by trust, merchant outcomes and business growth instead of operational automation alone.

Going Beyond Vanity Metrics

While every industry may define ROI on AI differently, enterprise technology companies are finding that the quality of AI systems matters as much as the quantity of deployments.

For MongoDB, accuracy has emerged as one of the strongest metrics. “The most important metric in the AI world is really around accuracy. The most successful applications are the ones you can trust,” said Benjamin Cefalo, CPO at MongoDB, an AI-native database platform used by the likes of Toyota and Visa.

The company believes better retrieval accuracy improves customer experience and reduces irrelevant context being sent to foundation models, lowering token consumption and improving application economics. Accuracy, therefore, is becoming both a technical and financial metric.

MongoDB is also seeing enterprises increasingly adopt databases as the memory layer for AI agents. Instead of evaluating AI by the number of deployed agents, organisations are beginning to assess whether these systems can reliably retain context, retrieve information in real time and continuously improve their responses.

As organisations scale AI deployments, investments in governance, compliance and data quality are becoming critical enablers of long-term business value. The importance of data quality is also becoming more evident. 

For instance at Axis Bank, real-time data infrastructure has enabled personalised customer experiences, improved operational efficiency, strengthened regulatory compliance and enhanced customer engagement, demonstrating that trusted data often delivers business value well before direct revenue gains become visible.

Taken together, these examples suggest organisations are moving beyond vanity metrics such as chatbot conversations or employee adoption rates. Instead, ROI in AI is increasingly being measured through indicators that directly influence business performance, including accuracy, trust, customer experience, governance, and decision quality.

The New Paradigm Of ROI 

Organisations have started evaluating AI not only by the costs it removes but also by the additional capacity it creates. A recent deployment of MiFiX.ai, an AI product engineering platform, at a large Indian public sector bank is a case in point. The AI platform automated a decades-old SWIFT reconciliation process, generating an estimated 500% ROI while reducing unreconciled funds by 95%. 

It also enabled employees to move from repetitive reconciliation work to customer-facing roles and created the operational headroom for the bank to launch new SWIFT-enabled remittance products, illustrating how AI can create entirely new business opportunities alongside efficiency gains.

A broader definition of ROI of AI is beginning to emerge across industries. For edtech companies, AI is increasing productivity without proportionately expanding teams. In healthcare, it is helping clinicians spend more time on expert judgement. D2C brands are measuring autonomous customer support alongside revenue growth. BFSI organisations are tracking merchant outcomes, fraud reduction and decision quality. Enterprise technology companies, meanwhile, are focusing on accuracy, governance and trusted data as the foundations for scaling AI reliably.

Different industries are calculating ROI differently, and that, perhaps, is the new maths behind AI ROI.

Top Stories From India & Around The World

• Indian AI Funding Jumps 4X In H1: Indian AI startups raised $676 Mn across 57 deals in H1 2026, more than four times the capital raised a year earlier. Investors attribute the surge to stronger startup quality, enterprise AI adoption and policy support, while arguing deeper capital pools are still needed to compete globally.
• C5i Files IPO Papers: Enterprise AI startup C5i has confidentially filed its draft IPO papers with SEBI, marking its second attempt to go public. The company is reportedly targeting to raise ₹1,000-1,200 Cr after growing FY25 operating revenue 26% to ₹545.3 Cr.
• Zeta Founder’s New Gig: Zeta founder Bhavin Turakhia has launched Neo, a self-funded enterprise AI startup backed by a $30 Mn personal investment. The platform combines AI assistants, agents, knowledge management and collaboration tools to embed AI across enterprise workflows.
• India’s Compute Crunch: Despite easing GPU shortages, demand for next-generation AI chips continues to outpace supply, stretching delivery timelines to months. Cloud providers are reserving capacity well in advance while startups optimise workloads as compute becomes a strategic competitive advantage.

The Weekly Buzz: Emergent’s AI Boardroom

For Emergent cofounder Mukund Jha, AI has already moved beyond being a chatbot or coding assistant into what he considers an executive operating system. Instead of spending the first hour of his day catching up on emails, Slack messages and meeting notes, he relies on an AI agent built on Emergent to prepare what he calls a ‘presidential briefing’. 

Every morning, the agent automatically pulls together follow-ups from previous days, unread emails, Slack conversations, pending action items for Jha and other updates into a single, contextual briefing that gives him a complete picture of the business before work begins.

The workflow goes a step further. Jha says he spends around 30 minutes every morning in conversation with his AI agent, discussing priorities, upcoming meetings, decisions and potential bottlenecks. Based on that discussion, the agent reorganises his schedule, reprioritises tasks and sends back an updated execution plan for the day. 

Rather than opening multiple AI tools or writing prompts throughout the day, the system functions as a personalised chief of staff that continuously remembers context and helps the Emergent cofounder make better decisions.

The workflow also reflects how AI-native companies are beginning to operate. Instead of treating AI as a productivity tool, they are increasingly building persistent agents that manage information, retain institutional memory and act as always-on operational partners for business leaders.

Startup In The Spotlight: SubVerse AI

Founded in 2023 by Tanmay Lad and Rishi Kumar, SubVerse AI is building AI voice agents and workflow automation systems that help BFSI enterprises automate collections, KYC, onboarding, renewals and customer support.

The startup’s platform is built on a three-layer architecture, including conversational AI, workflow execution and customer data orchestration, enabling enterprises to automate complete business processes instead of isolated interactions. SubVerse AI claims its platform integrates with various enterprise systems, including Salesforce, ServiceNow and HubSpot, allowing customers to deploy pilots within 30 days. 

Operating with enterprises such as Acko Insurance, the startup is positioning itself as an AI infrastructure layer that can execute operational workflows autonomously while reducing costs and improving efficiency. 

Looking ahead, SubVerse AI plans to expand its AI agent capabilities across more enterprise functions as organisations increasingly seek production-ready autonomous systems.

Prompt Of The Week

What prompts and hacks are CTOs, CEOs and cofounders using these days to streamline their work? 

Here’s the prompt Shubham Jindal, the director of AI at Harness, uses as a personal AI chief of staff to prioritise critical decisions, surface blockers and cut through the daily flood of Slack messages, emails, and meeting notes.

“I am an engineering leader overseeing AI platforms, agents, and infrastructure teams.

Review the Slack conversations, emails, and meeting notes from the last 12 hours. Filter out noise, status updates, and resolved discussions.

Identify:

• Decisions that require my input

• Team members blocked and waiting on me

• High-priority risks, escalations, or deadlines. Items where my intervention will have the highest impact

Present everything as a concise, prioritised briefing with recommended next actions, effort required, and ready-to-send responses where appropriate.”

Editor’s Note: Some prompts may need to be adjusted by users for best results or may not work as intended for certain users.

[Edited by Shishir Parasher]
[Creatives by Abhyam Gusai]

The post The New Math Behind ROI In AI appeared first on Inc42 Media.