Even if AI use cases look promising, they can still fail. The technology can generate output, automate tasks, and perform as designed, but the results aren’t always clear when leaders try to translate that work into cost savings or business improvements. This disconnect is evident in many organizations seeking to move beyond initial experimentation.
Companies are starting to realize that early advances in AI don’t always translate cleanly into production environments. Even use cases that work well in a controlled environment can be challenged when encountering real data, operational constraints, and the need to demonstrate measurable results. At the same time, many teams are still figuring out how to define and communicate ROI in a way that resonates with the business.
Piyush Saxena, senior vice president and global head of HCLTech Google business unit, pointed to a fundamental disconnect in a recent interview. This means that organizations are still measuring outcomes rather than performance. Generating summaries, automating tasks, or deploying agents may demonstrate functionality, but does not necessarily result in increased revenue, reduced costs, or improved operations.
This gap becomes even more pronounced when moving from proof of concept to production and when ROI is inconsistently defined. Without clear alignment between technical teams and business goals, initiatives will lose momentum before they can have a meaningful impact.
But some organizations are getting it right. They are becoming more disciplined in prioritizing use cases, aligning efforts to business key performance indicators (KPIs), and preparing operations for scale. In practice, this means focusing on the areas where AI can make a big difference in outcomes.
In the same discussion, Saxena highlights examples where AI is already delivering tangible results, from increasing supply chain visibility and reducing losses to increasing operational efficiency through automation. These are not isolated experiments. These are targeted applications that are directly related to business performance.
Bridging the gap between early success and lasting business impact
Even as organizations begin to see value, a second challenge quickly emerges: how to operationalize AI systems at scale. Mangesh Mulmule, vice president of HCLTech Google Business Unit, explains the fundamentally different operating environments. Traditional systems behave predictably. AI systems are not like that. They learn, adapt, and make probabilistic decisions based on constantly changing data. This creates a level of complexity that most organizations do not yet have the capacity to manage.
The operation of AI, especially agent AI, requires continuous monitoring. Models must be monitored, retrained, and managed in real time. Security risks grow as systems interact across environments. And as the lines between business and IT begin to blur, ownership becomes less clear. The implications are simple, but important. AI cannot be treated as a one-time implementation and requires an operational model.
Mulmule describes this as a coordinated effort across people, processes and technology. Governance needs to be built in from the beginning, rather than layered on later. Security should be applied continuously rather than periodically reviewed. And organizations need to rethink how they structure teams, define accountability, and measure performance over time.
For leaders, the path forward is becoming clearer. The focus is shifting from experimentation to execution. This means choosing fewer high-value use cases. Align early on measurable outcomes. Build the operational infrastructure needed to support AI in production.
This also means recognizing that scaling AI is not just a technical challenge. It’s also an organizational thing.
The two videos in this article explore these dynamics in more detail, from where AI is delivering real business value today to what it takes to operationalize AI at scale. Together, these provide 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 colleagues about your AI use cases.
