Generative AI tools like ChatGPT and Microsoft Copilot have accelerated innovation but also exposed enterprises to sensitive data, creating regulatory and operational risks that many organizations were unprepared to handle.
Abhishek Singh, Technology and Customer Service Business Executive at Amdocs, says the real challenge is organizational alignment, as AI adoption without clear policies and accountability can outpace governance and increase exposure.
Structured enterprise AI programs with centralized governance, operational boundaries, and human oversight allow companies to harness productivity gains safely while protecting data and maintaining trust.
Enterprise AI risk begins at the point of ingestion. The moment employees paste customer records, proprietary code, or board-level strategy into tools like ChatGPT and Microsoft Copilot without clear guardrails, governance gaps surface. What looks like productivity quickly exposes fractured data architecture, unclear ownership, and legacy controls that were never built for nondeterministic systems. The real divide is between organizations that treat AI as an operating model shift and build structured oversight early, and those that scramble to contain risk after exposure.
Abhishek Singh, a Technology and Customer Service Business Executive at Amdocs, a global leader in telecom software and services, leads enterprise operations worldwide. As a Service Business Leader overseeing millions in revenue and a Senior Member of both the IEEE and the Forbes Technology Council, Singh has insight in how rapidly innovation can outpace governance, revealing the operational and strategic challenges organizations face when adopting new technologies.
"The biggest risk is not AI outputs, it's the data you put into AI systems. If you don't protect that, you are exposing both your company and your customers," Singh says. The core issue is organizational alignment, not model quality. "People focus on hallucinations, but hallucinations don't trigger regulatory fine. Data leakage does. The enterprise problem is ingestion; what goes into the system and who is accountable for it."
Hiccups to high gear: In many enterprises, AI adoption began as a bottom-up experiment fueled by executive urgency. "When these tools first became widely available, we were going in all directions. Different teams were trying different models without a unified policy. It wasn't intentional risk-taking. It was lack of structure. This is where CIOs and CTOs must explain the strategy and define the path so that everyone in the organization understands," Singh explains.
The developer: Singh describes a project where a small team of engineers used advanced AI coding assistants, similar to Cursor-style tools, to accelerate software development. “A handful of engineers were producing output equivalent to a team of twenty. The productivity gains were undeniable, productivity without protection is exposure," he says.
Taming the beast: Layered governance works across multiple levels. "You need a governing body that provides the right path for the organization to follow. Otherwise, every business unit will interpret risk differently. Policy without education doesn’t work. AI can make decisions, but you have to establish thresholds that it cannot cross. Defining that threshold is critical," says Singh.
Organizations must move to boundary-based governance, defining safe operational zones where AI can function autonomously and clear thresholds where human oversight is required. This enables speed, productivity, and innovation without compromising security or accountability.
Sky-high: "When cloud came into the picture, most companies created dedicated governance units. Standards were centralized first. Then adoption expanded safely. The same pattern applies to AI. AI systems are nondeterministic. If you try to govern them like traditional software, you will fail. Boundaries, thresholds, and human oversight are the only way to ensure accountability while enabling autonomy," Singh adds. Traditional control-based governance, which assumes predictable behavior is insufficient.
Human checkpoints: "Our learning was that you have to have a human in the loop. You will know where you can hand things over to AI completely, but the ‘human in the loop’ will never go away. Leadership must always be able to demonstrate what controls were in place," Singh says. Oversight remains essential for regulatory compliance, customer trust, and operational accountability.
Enterprise AI maturity must be measured by the clarity of governance. By shifting from reactive risk management to structured AI governance, enterprises can protect sensitive data, ensure accountability, and safely scale autonomous systems. "The companies that will win with AI are not the ones experimenting the fastest. They are the ones building the clearest boundaries," Singh concludes.
