An AI capability is engineered at a GCC in India, funded through a global technology budget, deployed across European operations, and delivers its financial benefits in a North American business unit. At the quarterly review, who owns the return?
The GCC reports successful delivery. The business unit reports improved performance. Finance sees investment in one cost centre and benefits in another. The enterprise may have created real value, yet its measurement system cannot connect the dots.
This is the next challenge for enterprise AI: not merely proving that an initiative has positive ROI, but tracing how intelligence generates value across organizational boundaries, and how much additional value it enables when reused.
The Wrong Unit of Measurement
Traditional Return on Investment remains non-negotiable. However, a project-level business case captures only part of the economics when AI becomes a shared capability serving multiple products, functions, and geographies.
Return on Intelligence offers a complementary lens: the measurable value an enterprise realizes by converting data, models, domain expertise, and human judgment into better decisions, redesigned operations, new offerings, and capabilities that can be reused economically.
It is not a replacement financial ratio or a licence to count hypothetical future gains. The distinction is between value realized by the first deployment and the incremental value enabled by intelligence that the organization can deploy again.
Consider a multinational manufacturer whose India GCC builds an AI-enabled supplier intelligence platform. Its initial procurement team may reduce sourcing effort and identify supplier risks earlier. Those outcomes can support a conventional ROI calculation.
But the platform is then adapted for other markets. Procurement leaders use shared risk signals to avoid disruptions; finance teams gain better visibility into working capital; regional teams spend less time rebuilding data pipelines and decision logic. The investment's economic footprint now extends beyond the GCC's operating budget.
There are three returns to distinguish: direct gains from the first deployment, additional outcomes realized by other business units, and lower cost or faster time-to-value on subsequent deployments. They should be measured separately, not automatically added together. A saved analyst hour is not cash until capacity is reduced or demonstrably redeployed; an avoided disruption needs a defensible counterfactual; reuse must generate an actual benefit.
The Two-Ledger Approach to Intelligence Value
Enterprise leaders need two connected views rather than one expanding dashboard.
The Business Value Ledger captures what has been realized: attributable cost reductions, incremental contribution margin, working-capital improvement, and risk losses credibly avoided—net of engineering, inference, integration, oversight, and change costs. Each claim needs a baseline, a benefit owner, a geography, a time horizon, and a method for establishing attribution.
The Intelligence Capability Ledger records what makes future returns possible: reusable data products, domain-specific agents, validated decision patterns, evaluation assets, and the cost and time required to adapt them in another business unit. These are operating indicators, not accounting assets or financial gains in their own right.
The connection between the ledgers is decisive. When a shared model improves supplier forecasts, the capability ledger records where it is adopted. The value ledger records whether those decisions actually lower procurement costs or protect service continuity. Without that link, enterprises risk calling the same benefit a productivity saving, a working-capital gain, and an AI innovation dividend.
A concise 4I test can expose the evidence gap: Insight, did decisions improve? Integration, did end-to-end work change? Innovation, did a new product or value stream emerge? Institutionalization, was the intelligence reused with measurable incremental benefit? The first three explain how value is created; the fourth tests whether it can compound.
Why the GCC Mandate Must Change
India's GCC ecosystem, with 2,117 centres in FY2026 according to Nasscom–Zinnov, is increasingly positioned to own global products, platforms, and AI capabilities. Yet measuring GCCs primarily on headcount economics, delivery SLAs, or local productivity can obscure their contribution to enterprise results.
A GCC should not claim revenue uplift that a regional sales team delivered independently. Equally, its contribution should not disappear because the benefit is booked in a distant market. Value attribution requires joint ownership: business leaders are accountable for realized outcomes, GCC product and engineering leaders for the intelligence capability and its reuse, and finance for validating the bridge between them.
This also changes investment decisions. A platform with modest first-year savings may deserve funding if subsequent deployments have credible demand and materially lower marginal costs. Conversely, an elaborate shared asset with no committed adopters should not receive a premium merely because it is labelled strategic.
The Boardroom Question
Boards do not need another AI activity scorecard. They need visibility into where intelligence is created, where economic value is realized, who owns that value, and what can be reused without repeatedly paying the full cost of invention.
The shift from ROI to Return on Intelligence is therefore not a rejection of financial discipline. It is an expansion of accountability from a project to the enterprise.
The defining question is no longer only, "What did this AI project save?" It is also, "Where else did this intelligence create verified value, and what will it cost to make it useful again?"
The next AI advantage will belong to enterprises that can answer both.