Vinod Krishnan, CIO, Birlasoft explains how the CIO’s role is shifting from central builder to integrator, why AI governance must target outcome anomalies, and how trust overrides resistance.
As enterprises accelerate their AI transformation journeys, CIOs are increasingly being pushed beyond traditional infrastructure and application management roles into positions that demand business alignment, governance leadership, and organisational management. The rise of agentic AI, citizen development, and low-code platforms is reshaping how enterprises build, govern, and scale technology. Yet, while organisations are eager to experiment with AI, the larger challenge lies in operationalising it responsibly and embedding it into enterprise workflows without disrupting trust, accountability, or business continuity.
In this interaction with CIO&Leaders, Vinod Krishnan shares his perspective on how enterprise AI adoption is unfolding inside organisations. From the evolving role of the CIO and outcome-based AI governance to the realities of scaling AI beyond pilots, he explains why the biggest challenge in AI transformation is not technology itself, but the human resistance that accompanies every major operational change.
CIO&Leader: How do you see human oversight evolving in AI-led systems, as autonomous decision-making is still difficult in enterprises?
Vinod Krishnan: There is no enterprise decision-making with commercial or service implications that can happen entirely without human oversight. Organisations already operate on maker-checker models where accountability ultimately rests with a person who owns a cost centre or profit centre. Even if AI becomes the proposer or maker in many workflows, human approval remains essential.
I do not see organisations allowing AI systems to independently process purchases, payments, or customer-facing transactions without supervision. The concern is not only hallucination, but also accountability.
The way enterprises are approaching this is by creating boundaries and tolerances. If an AI system is managing cloud workloads, for example, there will still be guardrails around what it can and cannot do. Certain workloads may be marked untouchable, and spending thresholds may require escalation.
These are essentially the same operating principles we use for people. Some tasks are automatically approved, some require caution, and some require escalation. AI systems will eventually operate within similar governance pathways.
CIO&Leader: You have seen the evolution of enterprise IT closely over the years. How do you see the role of the CIO changing today?
Vinod Krishnan: Traditionally, the CIO in an organisation functioned as a black box. Business leaders understood IT delivered infrastructure, applications, and governance, but they did not necessarily understand how those outcomes were achieved.
The CIO’s responsibilities historically revolved around three pillars: infrastructure availability, application functionality, and governance/compliance. IT ensured employees had access to systems, business applications, networks, and data while maintaining operational standards and budgets.
What has changed significantly is the rise of citizen development.
During the low-code and no-code era, business users gained the ability to create applications independently. Earlier, users had to depend entirely on IT teams for every modification or workflow enhancement. Today, users can build functional tools themselves using AI-assisted coding platforms, no-code systems, or even Excel-based automations.
This creates both opportunity and risk.
If these user-built applications work well, organisations begin questioning the traditional role of centralised IT. But if they fail, business logic starts operating outside enterprise governance frameworks, creating security, compliance, and operational risks.
This fundamentally changes the CIO’s role.
The CIO can no longer operate only as the centralised builder of enterprise technology. The role is evolving into that of a coordinator, integrator, and scaler of business-led innovation.
The CIO can no longer remain only the builder of enterprise technology. The role is becoming that of a coordinator and integrator of business-led innovation.
Instead of spending months understanding requirements through meetings and documentation, CIO teams can now ask business users to build prototypes using AI tools or no-code platforms. The user expresses business logic through a working model, and IT then focuses on enterprise-grade integration, security, governance, and scalability.
This dramatically reduces implementation friction.
Earlier, both business and IT started from zero while trying to interpret requirements. Now, both sides begin midway through the journey because the business user already demonstrates the core logic.
The challenge for CIOs is whether they are confident enough to embrace this shift and say: “You have already solved part of the problem. We will now operationalise it properly.”
CIO&Leader: How are organisations approaching governance and responsible AI adoption, especially with agentic AI entering enterprise systems?
Vinod Krishnan: The activities being assigned to agentic AI today are activities already being performed by people. Organisations therefore already have governance structures that can be adapted for AI systems.
Governance is not about monitoring every intermediate step. It is about identifying anomalies in outcomes.
Governance in the age of agentic AI will not be about tracking every step a system takes. It will be about identifying anomalies in outcomes.
If a purchasing clerk typically releases purchase orders worth five million dollars weekly and suddenly releases twenty million dollars, governance mechanisms immediately detect the anomaly. The same logic applies to AI agents.
Organisations do not have the resources to shadow every action performed by an AI system. Instead, governance must focus on outputs and deviations from expected behaviour.
There will always be certain activities organisations do not want AI systems performing autonomously — firing employees, spending money, communicating with regulators, or sending reports to boards. These become negative-list activities requiring human supervision.
Other low-risk activities can tolerate varying degrees of automation and error.
That is how enterprise AI governance models will evolve; through clearly defined levels of acceptable autonomy.
CIO&Leader: When evaluating technology vendors and AI solution providers, what capabilities matter most to you?
Vinod Krishnan: The first requirement is basic competence. A partner must demonstrate the ability to execute what they are being asked to deliver.
The second is depth and diversity of experience. The more environments a partner has worked in, the stronger their generalised understanding becomes.
The third is long-term viability. One of the biggest risks with innovative technology vendors is sustainability. A company may have a brilliant product, but if it does not survive long enough, the customer inherits a major operational risk.
The fourth and most important capability is innovativeness, the ability to go beyond the stated brief.
There are three layers in enterprise demand: what the customer asks for, what the customer wants, and what the customer actually needs.
A user may request a button on a screen, but the real issue may be operational inefficiency in processing purchase orders. A strong partner must be able to interpret the underlying business problem rather than merely executing the stated request.
This is where domain knowledge becomes critical.
If a technology partner understands manufacturing terminology, operational metrics, assembly lines, or production realities, customers gain confidence that the partner genuinely understands the business context.
The ideal technology partnership is one where the customer explains a requirement, the partner interprets the business objective, identifies the actual need, aligns on outcomes, and then executes accordingly.
CIO&Leader: Many organisations successfully launch AI pilots but struggle to scale them. From your experience, what is the biggest barrier?
Vinod Krishnan: The biggest barrier is not infrastructure, data readiness, or ROI clarity. The real challenge is organisational resistance to change.
Whenever AI changes an operational process, people immediately become uncomfortable because responsibility still rests with them even when the system performs the task.
Take travel expense systems as an example. If an AI automatically scans receipts and fills expense forms, the first reaction is often anxiety rather than excitement. Employees immediately wonder what happens if the claim gets rejected or if the system makes a mistake.
This resistance is not unique to AI. It happened during ERP adoption, automation transitions, and every major enterprise technology shift.
Initially, users feel they are losing discretion and control. Over time, however, organisations calibrate themselves to the system and begin trusting it once consistency becomes visible.
Organisations initially resist AI through what I call organised resistance. People constantly ask: “How did the system generate this result?” or “Why is the AI recommending this outcome?”
Those questions are actually positive because they indicate engagement.
Leadership’s responsibility is to help people calibrate gradually. If the AI recommends one hundred and the user believes the answer should be forty, organisations should not abruptly reject the system. They should gradually build trust and familiarity.
AI adoption is fundamentally a trust-building exercise.
The challenge is that AI systems are less transparent than traditional enterprise software. In ERP systems, users could usually trace the logic step by step. AI often operates through pattern recognition that cannot always be fully explained.
That makes adoption psychologically harder.
Organisations do not naturally embrace disruption. Change creates uncertainty, and uncertainty creates resistance. If leadership fails to guide employees through that transition carefully, enterprises will reject AI transformation the same way the human body rejects a transplanted organ.
Organisations do not naturally embrace change. If AI is introduced without guiding people through that transition, the enterprise will reject it the same way the human body rejects a transplanted organ.
Successful AI adoption therefore requires gradual integration, trust-building, and organisational conditioning. Once acceptance happens, employees begin discovering entirely new efficiencies and use cases because the technology becomes part of the organisation’s operational muscle memory.