As businesses struggle to translate substantial AI investments into measurable returns, industry leaders are shifting focus from foundation models to comprehensive operational systems. During a recent discussion at VB Transform 2026, NTT DATA AIVista CEO Bratin Saha emphasized that the primary obstacle to AI adoption—often described as the 'last mile' problem—lies in bridging the gap between general-purpose models and the rigorous demands of regulated industry workflows.

According to VentureBeat, most enterprise AI initiatives falter due to inadequate integration, insufficient domain specialization, and a lack of clear governance. Rather than relying on simple model deployment, organizations are finding that success requires wrapping frontier models in proprietary data, institutional knowledge, and sophisticated guardrails. By creating an ensemble of systems that monitor and validate AI outputs, companies can better manage risks associated with complex sectors like multinational insurance, where models may otherwise struggle with specialized documentation and ambiguous input formats.

Saha argues that the most significant performance gains occur when enterprises shift their strategy from fine-tuning models to building robust, domain-specific systems around them. This approach allows organizations to steer AI behavior based on internal workflows and regulatory interpretations without relying on techniques like fine-tuning, which often rank lower in organizational priority. By prioritizing context-aware guardrails and proprietary data integration, firms can effectively transform capable foundation models into reliable, production-grade enterprise agents that deliver tangible business value.