At TechSparks 2026, one of the clearest messages to emerge from the enterprise AI conversation was that the biggest obstacle to adoption may not be the model at all. Michael Cronin of Couchbase said the real challenge sits deeper in the stack: fragmented data infrastructure that prevents organisations from feeding AI systems with clean, accessible and governed information at scale.
Data Before Models
Cronin's argument reflects a problem that many Indian enterprises are now confronting as they move from AI experimentation to deployment. While generative AI tools have captured executive attention, the operational reality inside most companies remains messy. Data is often spread across legacy databases, cloud applications, on-premise systems and departmental silos, making it difficult to create a single, reliable foundation for AI workloads.
That fragmentation matters because AI systems are only as effective as the data they can access. If information is inconsistent, delayed or trapped in incompatible systems, even the most advanced model can produce weak, incomplete or untrustworthy outputs. For enterprises, that translates into poor customer experiences, higher compliance risk and slower time to value.
Cronin's comments also point to a broader shift in how enterprise technology leaders are thinking about AI readiness. The early phase of AI adoption was dominated by model selection, prompt engineering and proof-of-concept demos. The next phase is more demanding. It requires data architecture that can support real-time retrieval, governance, security and scale across business functions.
The Infrastructure Gap
For startups and venture-backed companies building AI products, this is a critical market signal. Many founders have focused on model wrappers, workflow automation and user-facing applications, but enterprise buyers are increasingly asking harder questions about where the data lives, how it is synchronised and whether it can be trusted in production.
That is why infrastructure vendors such as Couchbase are positioning themselves as enablers of the AI stack rather than just database providers. The pitch is straightforward: before companies can deploy AI agents, copilots or decision systems, they need a data layer that can support low-latency access, flexible schemas and operational resilience. Without that, AI initiatives risk remaining pilots rather than becoming core business tools.
The issue is especially relevant in India, where large enterprises often operate across multiple legacy environments and rapidly expanding digital channels. Banks, retailers, telecom operators and consumer platforms may have abundant data, but abundance is not the same as readiness. Data quality, interoperability and governance remain persistent constraints, particularly when organisations try to move from analytics to generative AI.
Cronin's remarks also highlight why AI spending is increasingly shifting from flashy front-end applications to less visible but strategically important infrastructure. Investors and corporate buyers alike are beginning to recognise that the companies most likely to win in enterprise AI may not be those with the most impressive demos, but those that solve the plumbing problems underneath them.
Scaling Needs Discipline
The implication for executives is clear: AI strategy cannot be separated from data strategy. Companies that rush to deploy AI without fixing their underlying systems may see short-term excitement, but they are likely to encounter bottlenecks in accuracy, security and scale. In practical terms, that means modernising databases, reducing fragmentation, improving metadata management and creating governance frameworks that can support machine-driven workflows.
For the venture capital ecosystem, this creates both a challenge and an opportunity. The challenge is that many AI startups will need to prove they can integrate into complex enterprise environments rather than simply layer intelligence on top of them. The opportunity is that the market for data infrastructure, observability, governance and retrieval systems is likely to expand as enterprises realise that AI transformation begins long before the model is switched on.
TechSparks 2026 has featured many conversations about the future of AI, but Cronin's message stood out for its realism. In a market still captivated by model performance and product velocity, he argued that the decisive advantage may come from something less glamorous but far more durable: the ability to organise data well enough for AI to work reliably in the real world.
That framing is likely to resonate across India's startup and enterprise landscape. As companies race to adopt AI, the winners may be those that treat data infrastructure not as a back-office utility, but as the foundation of competitive advantage.