Google CTO Advisor Says Data Hygiene Alone Won't Deliver AI Value

The views and opinions expressed are those of Juan Antonio Peña González and do not reflect the views of any organization or publication.

The rapid pace of AI adoption is forcing a reckoning in the C-suite. As more employees turn to public AI, they risk feeding proprietary data into unsecured environments—a reality that's making AI governance a top-level conversation.

For an expert's take, we spoke with Juan Antonio Peña González, a Principal Enterprise Architect and CTO Advisor at Google. Drawing on over two decades of leadership in digital transformation, with roles including IT & Architecture Director at Chubb and Senior Information Technology Specialist at Boston Consulting Group, Peña González is deeply familiar with the source of the issue. From his perspective, many leaders are simply asking the wrong question.

According to Peña González, controlling data is now more important than any tactical debate over cloud versus on-premises infrastructure. Often, the reaction is to focus on "data hygiene," the expensive and time-consuming process of cleaning and structuring a company's vast reserves of unused, unstructured information. But the problem with this approach is that it fundamentally overlooks how modern AI actually operates, he explained.

"How do you manage the hygiene of the human mind? You can't," Peña González said. "You speak with a person to find common ground and determine if an idea aligns with business objectives. You don't try to clean up their thoughts beforehand."

• Past as prologue: Challenging the conventional view of this unstructured information as a liability, Peña González recast it as an untapped asset for innovation—a company’s collective memory. "We all have dark data in our own minds. It's the collection of our past experiences. How we innovate is by leveraging those past experiences and combining them with current trends," he explained. "The opportunity here is to leverage AI as a mind, using these data sources we aren't currently using to make them valuable."

• The art of the ask: Such a viewpoint encourages leaders to treat AI as a "digital worker" and focus less on data structure and more on the art of asking intelligent questions—a process enabled by the tedious but foundational work of data governance. "The best way to get information is to ask intelligent questions," Peña González said. "We're just going to have to think about how we question these new technologies and who can ask those questions."

A hybrid ecosystem is the most practical path forward, Peña González said. To address the governance problem, the model restricts sensitive processes to secure, local open-source models and allows public AI only for less sensitive tasks. Here, employees get access to sanctioned tools without compromising the company's most valuable assets.

• Secret sauce spill: As generalist AI becomes a commodity, leaders must look beyond existing use cases. Instead, the goal is to build competitive advantage through specialization, Peña González said. "This transformation is happening faster than any other technology. You cannot stop it. In cybersecurity, the weak spot is often the user. That's who will reveal your secret sauce to public AI models. Why not start now by giving them access to the models, where you'll be the one controlling how much data they can share and what they can share with it?"

For most companies, the advantage will come from companies building their own specialized models trained on unique datasets to solve specific, high-value problems. "You have just that good generalist worker, if you will, that can give you answers to generalist questions, but it's not going to think about the next step or in an evolution of anything. This is where companies are going to step in."

Ultimately, accountability for this strategy is clear: "This is not something that should be led only by IT," Peña González concluded. "It's something that should be led by the CEO and the other C-levels."