'AI is a culture and change transformation, and not just a technology shift': Satish H C, Executive Vice President at Infosys

Bengaluru: Scaling Artificial Intelligence (AI) requires a deliberate shift to modern, modular, and interoperable architectures, supported by data readiness, tech modernisation and governed access, said Satish H C, Executive Vice President & Chief Delivery Officer at Infosys. In an interaction with Uma Kannan of DH, he said AI is a giant leap in capability, but it is not system complete. He said innovation and governance are not trade-offs; they must reinforce each other.
Excerpts:

What does ‘responsible by design’ look like in practice when building and scaling AI across large enterprises?

Responsible by design is about architecting trust, transparency, and control into the system from day one—not retrofitting it later. It is about treating responsibility as a foundational design principle, not a compliance afterthought. Current enterprise landscapes built on deterministic principals operate in environments defined by scale, fragmentation, heterogeneity, and regulatory divergence to adhere to rule and regulation.

These environments are inherently complex, fragmented across platforms, heterogeneous in technology stacks, and subject to varying regulatory regimes. When AI is introduced into this landscape, the objective is not to replace deterministic systems, but to augment them with cognitive capability in a way that preserves reliability, predictability, and compliance.

In practice, this means wiring AI into the enterprise such that it can reason, learn, and adapt—without undermining the existing deterministic controls.

Artificial Intelligence must operate within clearly defined boundaries: knowing where it can act autonomously, where human judgment is required, and how decisions can be explained, audited, and overridden when needed. The cognitive power of Artificial
Intelligence should enhance decision‑making and efficiency, while deterministic systems continue to provide stability, traceability, and regulatory assurance.

What are the biggest challenges enterprises face when integrating AI with existing legacy infrastructure?

When you look at a typical enterprise landscape, it becomes clear why scaling Artificial Intelligence is hard. Legacy environments are fragmented, monolithic, and burdened by technical debt. Data is often siloed, inconsistently governed, and not readily usable for AI, making data readiness a foundational challenge. AI cannot simply be ‘plugged in’ to systems designed for deterministic, rule‑based workflows without addressing these underlying constraints. This is why many organisations get stuck in pilot paralysis—they can prove value in pockets but cannot industrialise it. Without modernised platforms, clean and accessible data, and standardised integration layers, AI initiatives struggle to move beyond experimentation.

Enterprises must invest in making data reliable, contextual, and AI‑ready, while securely integrating AI into existing systems of record. At the same time, they need to balance performance, compliance, cost, and explainability in a multi‑model world.

How do organisations balance the need for rapid AI innovation with governance, trust, and regulatory compliance?

Responsible AI frameworks provide structural guardrails, while business teams innovate within those boundaries. Organisations need to create safe playgrounds for experimentation, implementation, and innovation. Within these environments, governance and trust principles must be applied from the outset—covering data usage, model behaviour, security, and accountability—so innovation progresses within well‑defined boundaries.

Scaled engineering should make Artificial Intelligence fit for purpose for enterprise use. This means embedding responsible AI frameworks, observability, and lifecycle controls so that experimental success can be industrialised reliably. As autonomy increases, enterprises must embed safeguards against misuse—prompt injection, jailbreaks, model drift—while keeping humans firmly in the governance loop.

Understanding, auditing, and overriding AI decisions is not optional; it is foundational for trust and scale. When innovation is enabled through structured experimentation and brought to production through governed, scaled engineering, enterprises can move fast without compromising compliance or confidence.

What key lessons have emerged from deploying AI at scale that enterprises often underestimate at the start?

The biggest lesson is that Artificial Intelligence is first and foremost a culture and change transformation, and not just a technology shift. Simply introducing new systems without preparing people for how to work with them creates friction.

Scaling AI requires training teams on new ways of interacting with systems, new workflows, and new decision flows—this is a fundamental cultural shift.

 As AI is embedded into enterprise operations, the way work gets done changes. Teams must reimagine processes rather than replicate old ways of working with new tools.

This introduces a new paradigm—new flows, new roles, and new expectations—where humans move from execution to oversight, judgment, and exception handling. Skills therefore need to evolve alongside systems, combining domain expertise with data literacy, AI fluency, and systems thinking. Organisations that invest early in change management, skills evolution, and responsible design are the ones that succeed in scaling Artificial Intelligence with confidence.