AI’s Biggest Barrier Is Fragmented Data

Four years after the launch of ChatGPT ignited the AI boom, businesses are entering what has been dubbed the “year of AI ROI.” Many now consider themselves AI-ready, pointing to successful chatbot and copilot rollouts as evidence.

But experimenting with AI is not the same as embedding it across complex business processes. The author, Nasuni’s chief product officer, argues that the biggest obstacle is often below the model layer: data infrastructure.

Why fragmented data blocks AI projects

Disjointed file environments, inconsistent governance and information spread across multiple repositories were manageable when data was accessed sporadically by human employees. People could compensate when information lacked context or was stored somewhere unexpected.

Recommended reading

Why the SaaS apocalypse may be overstated

AI systems cannot work around those gaps as easily. They require fast, reliable and continuous access to information, along with clear governance and sufficient context. When data is difficult to locate or inconsistently accessible, an enterprise may technically possess the information while being unable to use it effectively.

Traditional file environments were designed around separate locations, team-managed access and operational compromises. Those inefficiencies were frustrating for employees, but they did not necessarily stop work. AI makes them a structural limitation.

Early AI wins can hide deeper problems

Pressure to show progress is encouraging organizations to deploy new AI tools quickly. Chatbots and copilots offer a relatively low barrier to entry and can produce visible, tangible results.

The risk is that these early successes create overconfidence about an organization’s broader readiness to scale AI. A useful copilot rollout does not prove that the underlying data is accessible, governed or resilient enough for larger systems. The same organizations may also be struggling with data recovery after cyber incidents—another warning that their foundations remain weak.

As businesses move toward major agentic AI projects, they can encounter delays, unresolved questions about return on investment and failed implementations. The urgency to deploy AI can therefore slow longer-term progress when risk management and data management work are postponed.

Treat enterprise data as a strategic asset

Most businesses do not lack data. Their problem is that unstructured information—the location of many of their data assets—remains fragmented and difficult for IT teams to manage. Data is still often treated as an operational concern involving capacity planning, refresh cycles and expansion rather than as a strategic resource.

AI changes that calculation. Systems operate continuously, drawing on information created today, yesterday and sometimes decades ago. Meaningful results depend on trusted data, rich context, consistent access and clear governance.

The author says C-suites should shift their focus from steady-state storage capacity to data utility: centralized environments with fewer systems, where information can be used regardless of where it is stored. Reducing fragmentation and creating a unified view of enterprise data can make it easier to move from isolated AI successes to deployments that improve productivity and efficiency.

Chatbot or copilot adoption alone is not a measure of AI readiness. That readiness depends on whether the data behind those tools is accessible, secure and fit for purpose at scale. Organizations that invest in the data layer and governance before expanding their AI ambitions will be better positioned to turn early experiments into sustained value.