For twenty years, the answer to who was responsible for the CIO’s mistakes was crystal clear. Technology teams built the tools, business teams made the decisions, and the business took accountability. However, AI’s advent has changed the rules of the game. If a machine-learned model makes an inappropriate decision (such as denying a valid loan application, shipping products to the wrong address, or fabricating factoids that breach compliance), this age-old separation between the people who built it and the people who are responsible no longer holds true.
CIONOW spoke to four senior technology leaders to seek their perspectives on accountability.
The starting point, according to Golok Kumar Simli, President-Technology & Innovation at BLS International, is that technology has stopped being something that happens downstream of business decisions.
“In the AI era, technology is no longer a back-office function. Business sits in the driving seat, and technology is welded into every process: marketing, operations, HR, finance, customer experience,” he says. “CIOs and CTOs are now at the centre of the boardroom conversation, not behind it. You cannot launch new business streams without technology; you cannot shut them down safely without it either.”
This precisely complicates the accountability question. When technology sat at the edges of the business, failure was easy to trace to a function head. When technology is welded into every process, failure has no single owner by default. Accountability must be assigned deliberately.
Shared Blame, But Not Equal Duty
Every technology leader we spoke to echoed the same sentiment: accountability for AI must be shared. None of them believe it is fair or functional to pin failure entirely on IT.
Dr Tejasvi Addagada, Head of AI Governance at HDFC Bank, says, “If something goes wrong, it’s usually the functional head or the business head who is held accountable. Technology is there to enable business and bring business agility. The business owns the balance sheet; technology doesn’t.” IT’s job, in his view, is to guarantee that internal controls are sufficient, and not to absorb blame for decisions made elsewhere.
Balaji Raghunathan, Data & AI Engineering Business Unit Leader at Sigmoid, says, “AI ownership must be shared, but the exact model depends on enterprise maturity. The technology organisation should own the platform, infrastructure, reliability and security. Business functions should own the quality of domain data, the use case, the operating process and the business outcome.”
In a federated data setup, he says, the function that generates the data — supply chain, sales, marketing — owns its quality, while the CIO’s team owns the horizontal platform it runs on. “This creates clearer accountability than asking IT to carry responsibility for data it does not create or use daily.”
But shared doesn’t mean equal. Simli points to a duty that sits uniquely with the technology leader. “The technology leader does carry a special obligation: risk assessment, risk scoring, continuous evaluation, and explainability. That ‘human layer’ is ours. We are often the only ones who truly understand how the model thinks, and therefore the only ones who can properly explain its behaviour and limitations,” he says. Explainability, in other words, isn’t a shared responsibility; it’s a CIO responsibility because no one else in the room is positioned to carry it.
Communication is the Real Control
If there’s a second theme running through these conversations, it’s that most AI failures aren’t really technology failures. They are communication failures.
Arun Goyal, CIO at Sir Ganga Ram Hospital, is blunt about where projects actually break down. He says, “A CIO alone cannot implement any digital perspective unless he communicates and discusses with the stakeholders. If you do this in silos, you create siloed projects destined to fail.”
His bigger concern isn’t model accuracy; it’s expectation management. Stakeholders, he argues, chronically overestimate what AI can do unsupervised. “They should not think that once AI is there, they will sit quietly and everything will be done by the AI,” he says.
The CIO’s job, as he sees it, is as much about educating business stakeholders on what a technology can and cannot do as it is about building the technology itself. Get that shared understanding right, he suggests, and “it will never fail” because everyone is accountable to a version of success they actually agreed on.
The Real Test isn’t Adoption, it’s Readiness
Balaji believes it is not whether to adopt AI. It’s whether the organization has done the unglamorous work first: “the data, architecture, governance and business accountability needed to make AI dependable at scale. Once AI moves out of pilot mode, it becomes a CIO-level concern by default, not because CIOs asked for the exposure, but because uptime, integration, data reliability, cybersecurity and scale all report to them regardless of who signed off on the use case,” he says.
The accountability conversation, then, isn’t really about assigning blame after the fact. It’s a governance decision that has to be made before the model goes live — who owns the data, who owns the outcome, who owns the explanation when a board member asks “why did it do that?”
Enterprises that wait until something breaks to answer that question will find, as these four leaders suggest, that the answer was never going to be simple and by then, it’s too late to design it properly.
Conclusion
Business heads own the outcome and the balance-sheet impact of AI use cases. CIOs own the platform, security, reliability, and critically the ability to explain how models work and where they can fail. That split must be documented before go‑live, with clear owners for data quality, monitoring, and incident response. Treat accountability as a design decision, not a post‑incident debate.
