Since the widespread emergence of generative AI tools in late 2022, artificial intelligence has become a central focus for most organizations.

Adoption has been rapid. A vast majority of organizations are now using generative AI in some form, often through off-the-shelf tools designed to generate content, write code, or summarize information.

Think of popular tools such as Microsoft 365 Co-pilot, Otter.ai, Jasper, Git Hub Co-pilot, or the thousands of other AI tools on the market that are in use by organizations today.

This level of activity suggests meaningful progress. However, the reality is more nuanced.

Industry data suggests that many organizations are not gaining measurable or consistent value from their AI initiatives, despite widespread adoption. Those who are deploying off the shelf generative AI tools are typically seeing the least measurable impact.

Those who don’t think of AI as a tool, but as a set of capabilities are often more mature in their deployments and are more likely to measure actual impact. In other words, for many, what appears to be progress with AI is just activity because there are no foundations in data, skills, governance or operational know-how to move past. It is important for leaders to understand the difference.

The challenge

Part of the challenge stems from a disconnect between how AI is experienced and what is required to deploy it effectively in a business context. Consumer-facing AI tools are designed to be intuitive and immediate. They create the impression that AI can be deployed quickly and at relatively low cost.

In contrast, enterprise AI, AI that often uses your company’s data depends on a different set of capabilities. It requires access to reliable and well-governed data, integration with existing systems, alignment with business processes, technical skills, and oversight mechanisms that ensure trust and accountability.

This distinction is reflected in adoption patterns. While the majority of organizations are using off-the-shelf generative AI tools, far fewer have progressed to more advanced implementations. Typically, as complexity increases, adoption declines. Many organizations are still operating at an early stage, even as expectations, which are often shaped at the leadership level, continue to rise.

When AI Stalls

When AI initiatives stall, the underlying causes are typically not technical. The models themselves are increasingly capable, and the tooling ecosystem continues to evolve rapidly. Instead, the breakdown tends to occur at the organizational level.

Across industries, similar patterns emerge: data is in silos and organizations are not convinced of its integrity, platforms are not fully integrated, ownership of AI initiatives is unclear, and governance structures are either incomplete or not consistently applied.

In many cases, leaders are asking their organizations to introduce AI into environments that were not designed to support it. These are not isolated technical gaps; they are the result of decisions about data, governance, and investment that sit at the leadership level.

Organizations that are achieving more consistent outcomes tend to approach AI differently. Rather than treating it as a set of tools or isolated projects, they treat it as an enterprise capability that must be developed deliberately over time.

This includes investing in data foundations that integrate structured and unstructured data, establishing governance frameworks early, and aligning business and technical teams around clearly defined use cases. The role of enterprise data is particularly important. Without this grounding, AI remains limited to generic outputs and cannot reliably support business decisions or workflows.

Path to value

The path to value is less about initial adoption and more about depth of implementation. Organizations often begin with productivity-focused use cases using off-the-shelf tools, but meaningful impact tends to emerge as they incorporate enterprise data, integrate AI into workflows, and establish the processes required to operate these systems reliably. The difference is not simply technical execution; it reflects deliberate choices about where to invest and how to scale.

At this stage, the role of leadership becomes central. In many organizations, AI is still treated as a technical initiative and delegated accordingly. That approach is one of the primary reasons initiatives stall. AI requires coordinated investment across data, governance, architecture, and operating models—areas that sit squarely within executive responsibility.

Responsible leadership

Leaders ultimately determine whether AI remains a series of experiments or becomes an enterprise capability. This includes setting a clear strategic direction, funding foundational capabilities rather than isolated projects, and ensuring that AI initiatives are tied to measurable business outcomes. It also requires developing enough understanding to evaluate trade-offs, assess risk, and ask more precise questions about scalability and value.

The emergence of agentic AI brings these issues into sharper focus. Unlike earlier forms of AI, agentic systems are designed not only to generate outputs but to take action within workflows. This introduces new opportunities for efficiency and automation, but it also increases the importance of governance, control, and reliability. Despite growing interest, very few organizations currently have multi-agent systems in production. This gap underscores how early most organizations are and how significant the distance remains between experimentation and true operational capability.

The next phase of AI adoption will likely be defined less by experimentation and more by execution. Organizations that succeed will be those that move beyond isolated initiatives and focus on building integrated, enterprise-level capabilities. This will require coordination across data, technology, governance, and organizational processes, supported by sustained executive focus.

AI itself is not the limiting factor. The constraint is whether leadership is willing to build the organizational capability required to support it.

We feature the best Robotic Process Automation (RPA) software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Founder of the AI Foundations Group.