- Telcos are confident about data visibility and access, but AI execution is lagging
- Agentic AI raises the stakes for data governance
- Data quality becomes critical when humans aren’t in the loop
It’s one thing to experiment with AI, and another thing entirely to deploy it at scale. This hard truth is something enterprises across the board are learning in real time, but it turns out there’s a key to making it all work: data.
Cloudera recently surveyed 1,300 global IT leaders, including telcos, about their AI adoption and data strategies. The results, compiled in its Data Readiness Index, showed 89% of telecoms companies said they have complete visibility into where all of their organization’s data resides and 84% said they can access their data at any time regardless of format or storage location.
But “while respondents from telecommunications organizations reported higher confidence in visibility and accessibility, that success hasn’t totally translated into operational success,” the report noted.
So, what’s the problem? Well, 60% of telecommunications respondents said infrastructure performance hinders their operational initiatives. But even more notably, only a third of telecoms said all of their data was fully governed.
That governance gap is about to become a major liability in the world of agentic AI.
“Even though data can be accessible, oftentimes they don’t have the right layer of data governance in place” that allows them to both see the data and trust it in the hands of an agent, Cloudera's CTO Sergio Gago told Fierce.
Take OpenClaw, for example. Sure, its capabilities come with a certain “wow factor,” Gago said. But in order for it to be truly useful it has to have access to a telco or enterprise’s entire data estate. Without governance, that doesn’t happen.
“You data is going to scattered all around. It needs to be governed because otherwise you cannot trust either the access or the data itself. If you cannot trust it, you cannot use the agent and then it becomes useless,” he said. “Then you become one of those enterprises who say they’re fully into AI but all they’re doing is using CoPilot to summarize emails. That’s nice but it doesn’t cut it on true competitive advantage.”
Quality over quantity
But it’s not just about governance – Gartner has noted that poor data quality can negatively impact AI’s performance by 30%. And data quality is going to become even more important in an agentic world.
Most of us are probably familiar with the phrase “garbage in, garbage out.” In the early days of AI, there was always a human in the loop to take out the trash. When agents start operating independently, however, that won’t be the case, making data quality of utmost importance.
Gago argued that a single layer of data governance is necessary to dictate which agents can access what data and attest that the data being used meets certain quality standards.
“When you start to do automation and agentic AI you don’t have that domain knowledge anymore. So, now healing and data quality becomes a fundamental element of your policy,” Gago explained. “An agent needs to understand that if a given data point passes a certain threshold of quality loss, the agent should not be able to use that one either for its training, for its reasoning or for the final output.”
Cloudera isn’t the only one to have highlighted the AI data problem. Fierce Network Research last year shined a spotlight on the need to break down data siloes and democratize data access to create modern pipelines for AI environments.
Meanwhile, Gartner noted strong governance can increase AI success rates while reducing compliance risks. Audit and consulting firm PwC similarly stressed the importance of data governance for AI.
“AI raises questions about data lineage, sources and usage rights, especially when using data in the context of an AI solution,” the company wrote. “Data governance is no longer just a guardrail, it’s a launch pad. Without strong governance, AI systems may produce unreliable results and increase risk.”
PwC recommended elevating data governance to a board-level priority, modernizing architecture and tooling to enable better governance and rationalizing data sources as first steps in the right direction.