OPINION | From AI pilots to enterprise platform transformation

AI remains a work in progress, largely in pilot mode.

In terms of its impact on work, life and societies, Artificial Intelligence (AI) ranks alongside electricity, railways, printing presses, steam engines and automobiles — perhaps even above them. AI is among the top three strategic priorities for nearly three-quarters of enterprises globally, with one in five naming it their top agenda item. In 2025 alone, companies increased AI investments by roughly 80%. Across multiple domains, AI systems now outperform human benchmarks. AI is rapidly evolving from generative assistance to agentic autonomy — systems that can reason, orchestrate workflows and act with limited supervision.

Yet, for all its disruptive promise, only a few enterprises are meaningfully scaling AI across the organisation. Most remain stuck in experimentation — running pilots, launching copilots and celebrating isolated productivity gains — even as the 20% that are scaling are seeing meaningful results. For India Inc., AI remains a work in progress, largely in pilot mode.

Inside most enterprises, AI is confined to limited use cases: a chatbot in customer service, a code assistant for developers, or a dashboard that provides summaries — almost as if to reconfirm that AI delivers value.

This is the ‘micro-productivity trap’. Companies capture incremental gains — a few percentage points of efficiency in procurement, faster drafting of emails, marginally improved analytics — but fail to rewire processes end-to-end. They automate parts of processes instead of redesigning them across the organisation. This is where the real promise lies.

AI is treated as a technology add-on rather than as a lever for business transformation — one that requires not just changes in tech stacks but shifts in behaviours and operating models. As a result, pilots proliferate and pause without progressing to scale. The return on investment (ROI) remains ambiguous. Organisational fatigue sets in. The outcome: experimentation without enterprise impact.

Redefining the Value Case

AI’s true promise is not limited to cost savings. Across major functions such as sales, operations, supply chain and finance, the full productivity impact ranges from 10% to 30%. The larger goal should be structural competitive advantage: faster decision cycles, enhanced throughput, superior customer engagement and new revenue models.

Scaling AI requires an organisational shift. It demands visible C-suite sponsorship, deliberate change management efforts and meaningful employee reskilling. Companies that focus rigorously on delivery rather than experimentation are two to two-and-a-half times more likely to sustain results. Those that do not risk becoming the ‘corporate Sisyphus’, repeatedly pushing AI initiatives uphill only to watch them roll back down.

To make outcomes stick, enterprises must change how they work when deploying AI and reimagine how they operate. One approach is the ‘Lab and Crowd’ model. Here, a central AI lab sets the standards and builds the core systems, while teams such as marketing, finance, operations and HR develop practical applications within their functions. The ‘Lab’ ensures AI is reliable, safe and consistent; the ‘Crowd’ ensures it is useful and drives ROI for the enterprise.

Building the Platform

The next phase of AI will not be powered by isolated copilots. It will be driven by enterprise-grade, agentic platforms. These systems go far beyond large language models. They require orchestration layers, memory, tool access, protocol libraries, observability and security frameworks. Agents will communicate not only with other agents and technologies but also operate within human environments, requiring tools and guardrails to safely integrate human and agent behaviours.

In the near term, agentic AI will coexist with legacy enterprise architectures. Over time, however, organisations will need foundational intelligence layers that are event-driven, secure, interoperable and deeply integrated with enterprise data and context.

This is where many companies falter. They underestimate the importance of data architecture, governance and FinOps discipline. Scaling exposes critical gaps in operationalising AI, transforming a technical project into a complex cultural and structural challenge. Key hurdles include the absence of an AI operating model, persistent data silos and the need to shift from measuring outputs to measuring tangible ROI.

AI stack redesign must begin with business architecture and data advantage, not technology alone. Domain-centric ownership of agent development ensures that knowledge resides where value is created. Governance must evolve to ensure explainability, risk control and compliance, especially in regulated sectors such as banking, telecoms and healthcare, which are central to India’s economy.

In some parts of the world, labour costs and trade dependency dampen immediate AI-driven ROI. This is where India’s scale and digital public infrastructure create fertile ground for platform-led transformation. This is not merely a technology moment; it is a leadership moment.

The path across the pilot-to-platform chasm is clear. It requires bold ambition, disciplined execution, ground-up redesign and change-oriented leadership. The alternative is to remain trapped in micro-productivity — celebrating incremental gains while competitors redesign their businesses from the ground up. In the age of AI, structural advantage will not come from the best pilot. It will come from the best platform.

(Arpan Shethis a Partner in Bain & Company’s Mumbai and Washington, DC, offices.)

He is the head of Bain’s global Innovation & Design practice, and also leads their Private Equity and Alternative Investor practice in India.

Views are personal, and do not represent the stance of this publication.