By Prasad Rai, CEO, DAAS LABS
Every boardroom in India now carries an AI slide, yet remarkably few carry a data balance sheet. That imbalance, far more than the choice between rival foundation models, will determine which enterprises convert the current wave of enthusiasm into durable earnings. Models are commoditising along a pricing curve that bends only downward, and whatever one firm licenses today its competitor can license tomorrow at a lower price. Proprietary, well-governed, decision-ready data remains the one asset in the AI stack that cannot be downloaded.
Capital markets appear to have reached this conclusion ahead of many operating plans. Nasscom and BCG project that India’s AI market will reach 17 to 22 billion dollars by 2027, compounding at 25 to 35 per cent a year.
Growth of that order attracts scrutiny along with capital, and the questions asked in earnings calls and diligence rooms have shifted accordingly, away from which pilots a company is running and toward what those pilots have actually done to margin, working capital and risk exposure. Enterprise value is quietly repricing around demonstrated data maturity rather than announced intent, which makes the next set of numbers uncomfortable reading.
From Proof of Concept to Profit and Loss
Experimentation in India is abundant while production remains scarce. IDC found that 76 per cent of Indian enterprises were already pursuing generative AI proofs of concept or had firm investment plans, with national AI spending on course to touch 6 billion dollars by 2027. Yet as recently as June 2026, Nasscom observed that only about a quarter of technology services companies had moved AI experiments into production, and it named data readiness among the capabilities that now separate value creation from noise.
The diagnosis behind that shortfall is rarely the model itself. Pilots die in the plumbing, where customer records sit fragmented across a dozen legacy systems, where definitions of revenue differ from one business unit to the next, and where lineage cannot be traced on the day a regulator or an auditor asks for it.
An algorithm trained on such foundations does not fail loudly in a demonstration. It fails quietly in production, where every wrong answer carries a real cost. Enterprises that treat data as genuine infrastructure, engineered and governed the way a nation maintains its highways and power grids, sidestep the trap altogether. In India, that analogy is no longer a figure of speech.
Public Rails, Private Freight
India has spent a decade demonstrating that shared digital infrastructure is compounding. Aadhaar, UPI and the wider India Stack turned identity and payments into public rails on which private innovation now runs at population scale. The IndiaAI Mission and the Digital Personal Data Protection Act extend that same logic to computation and to data itself, with one expanding capacity and the other hardening trust. The state, in effect, is laying the roadbed.
What no national mission can supply is the freight. The quality, standardisation and governance of each enterprise’s own data estate remain a private responsibility and an unforgiving one under a consent-driven privacy regime. Seen clearly, governance is not compliance overhead but the very mechanism that makes data usable at speed, since the lineage that satisfies a regulator is the same lineage that allows a credit model or a demand forecast to be trusted on a Monday morning. Banks, retailers, logistics networks and manufacturers that standardise early will find each successive AI initiative cheaper and faster than the last, and that compounding effect is precisely where individual firms and their leaders should now direct their attention.
An Agenda for the Next Board Meeting
The practical agenda is one page long. It begins with an honest maturity assessment, because ambition without a baseline is theatre. It proceeds to a unified data fabric spanning cloud and legacy estates so that information moves with permission and provenance attached and to ownership written into the operating rhythm, because data quality decays the moment it becomes nobody’s job. It ends with culture, the hardest line item of all, in which decisions default to evidence and data and AI become how the organisation thinks rather than something it merely buys.
None of this photographs well at a launch event, yet all of it determines who wins. The AI-first organisation will not be the one that selects the cleverest model in 2026 but the one whose information already flows like power through a grid, priced, governed and available wherever a decision is being made. The race is run in the unglamorous mile where data meets intelligence. The model is merely the final hundred metres, and by then the outcome is usually already decided.