From AI experimentation to operational impact: What leaders need to get right

An AI use case may look promising and yet still fail. The technology can generate output, automate tasks, and perform exactly as designed, but when leaders try to connect that work to cost reduction or operational improvement, the results aren’t always clear. That disconnect is showing up for many organizations as they try to move beyond early experimentation.

Enterprises are finding that early progress with AI does not always carry cleanly into production. Use cases that perform well in controlled environments can struggle once they encounter real data, operational constraints, and the need to demonstrate measurable outcomes. At the same time, many teams are still working through how to define and communicate ROI in a way that resonates with the business.

In a recent interview, Piyush Saxena, SVP and global head of the HCLTech Google Business Unit, pointed to a fundamental disconnect: Organizations are still measuring outputs instead of outcomes. Generating summaries, automating tasks, or deploying agents may demonstrate capability, but they do not necessarily translate into revenue growth, cost reduction, or operational improvement. 

That gap becomes even more pronounced in moving from proof of concept into production and when ROI definitions remain inconsistent. Without clear alignment between technical teams and business objectives, initiatives lose momentum before they deliver meaningful impact. 

Yet some organizations are getting it right. They are more disciplined about prioritizing use cases, aligning initiatives with business key performance indicators (KPIs), and preparing operationally for scale. In practice, that means focusing on where AI can materially change outcomes.

In the same discussion, Saxena highlights examples where AI is already delivering measurable results, from improving supply chain visibility and reducing losses to increasing operational efficiency through automation. These are not isolated experiments. They are targeted applications tied directly to business performance. 

Closing the gap between early success and sustained business impact

Even when organizations begin to see value, a second challenge quickly emerges: how to operate AI systems at scale. Mangesh Mulmule, vice president, HCLTech Google Business Unit, describes a fundamentally different operating environment. Traditional systems behave predictably; AI systems do not. They learn, adapt, and make probabilistic decisions based on constantly changing data. That introduces a level of complexity most organizations are not yet equipped to manage.

Operating AI, particularly agentic AI, requires continuous oversight. Models must be monitored, retrained, and governed in real time. Security risks expand as systems interact across environments. And ownership becomes less clear as the boundaries between business and IT begin to blur. The implication is straightforward but significant. AI cannot be treated as a one-time deployment, and it requires an operating model.

Mulmule describes this as a coordinated effort across people, process, and technology. Governance must be embedded from the start, not layered on later. Security must be continuously enforced, not periodically reviewed. And organizations must rethink how teams are structured, how accountability is defined, and how performance is measured over time. 

For leaders, the path forward is becoming clearer. The focus is shifting away from experimentation and toward execution. That means selecting fewer, higher-value use cases; aligning on measurable outcomes early; and building the operational foundation required to support AI in production.

It also means recognizing that scaling AI is not just a technical challenge. It is also an organizational one. 

The two videos within this article further explore these dynamics — from where AI is delivering real business value today to what it takes to operationalize it at scale. Together, they offer a practical view of what is changing and what leaders need to do next.

Visit booth 1901 at Google Cloud Next to chat with HCLTech, Google Cloud, and your peers about AI use cases.