As part of the Government AI Campus, Apolitical is publishing a series of articles featuring leaders from the 2026 Government AI 100, a list recognising public servants around the world who are pioneering AI adoption, capacity building and regulation.

Apolitical spoke with Patricio Moyano Peña, who oversees a 175-person team focused on redesigning the justice service through people-centred innovation, responsible technology adoption and improved digital experiences. Buenos Aires Province has a population of 17.5 million. Its Public Prosecutor's Office is composed of 8,700 public officials, encompassing the functions of prosecution, public defence and guardianship, and intervenes in approximately one million criminal proceedings annually.

What problem were you most focused on when you started your role, and how has that focus changed as AI capabilities have evolved?

In 2017, our main problem wasn’t technology, it was coordination. People had ideas, but those ideas didn’t spread across the organisation. We had teams building similar solutions in parallel, and information was mainly paper-based and sometimes not recorded.

So our focus at first was governance and finding a way to scale innovation. To address this, we created a unit to centralise and scale innovation. Before the pandemic, we had to convince teams that digital transformation mattered. They saw AI as ‘sci-fi’ and not useful for daily work, so our main effort was cultural. During the pandemic, expectations shifted, and users wanted more intuitive, seamless and faster tools.

With the rise of GenAI, we started to experience coordination challenges again, because users felt empowered to experiment with these tools outside our institutional framework. So it became a cycle: first governance and coordination, and now again adapting this technology in a fast but responsible way.

What were the hardest debates when turning ethical principles into an operational framework?

Our hardest debate wasn’t about values, everyone agreed we should be responsible. The real debate emerged when we tried to put those principles into practice. Responsibility means more friction. We have to add steps, create clearer accountability and sometimes limit user autonomy. This creates tension between short-term productivity and long-term institutional risks.

< “The benefits of GenAI are immediate while the risks are sometimes invisible and delayed.”

Users may accept the trade-off because they want to be more productive, but we need to manage that safely. Confidentiality is a clear example. Users wanted to upload sensitive information into external systems. We had to explain the risks and provide safer alternatives that allowed them to do the same work.

Can you share an example of how one of these systems changes the day-to-day of public servants?

Identity verification in criminal cases is critical. If you miss prior records, you can make serious mistakes. In 2017, this process was highly manual and paper-based, taking weeks or even months. This could potentially lead to errors.

We built a multibiometric identification system that combines fingerprints and facial images to suggest matches across databases. It does not make decisions; a trained forensic expert reviews every case. The impact has been significant. Processes that used to take weeks or months are now solved within a day, and we have processed over half a million cases.

What have you learned about building AI literacy among non-technical professionals?

One key lesson is understanding how users interact with AI. Justice professionals work under pressure and cognitive load, and GenAI can amplify biases like anchoring or fatigue. Having a ‘human in the loop’ is not enough on its own. People need training, time to review outputs and confidence to disagree with the system.

A second lesson is that good tools are not enough. We held co-design workshops with around 200 users, which helped align expectations and build solutions with empathy. This showed that technical teams felt undervalued and users felt unheard. Co-design helped address that. We learned that the user must be at the centre, recognising human limitations while building solutions together.

What have you learned about balancing innovation with safety and ethics?

Sustainable innovation in the public sector requires governance, culture and technology working together. Governance means clear frameworks grounded in legal principles. Culture means reinforcing that responsibility cannot be delegated and AI should support, not replace, judgement. Technology means building safeguards by design, such as access controls, audit logs and rules about what people can input. We also differentiate between tools for low-risk and sensitive tasks, offering different environments for each. We started with pilots and then scaled, continuously improving both infrastructure and models.

What tends to break when moving from pilots to scale?

What breaks first is the pilot itself. With a small, motivated team, things work, even without a clear problem definition. At scale, that fails. The second issue is the underlying system. Scaling isn’t just deploying technology; it’s understanding context.

“If you have poor quality data, you’ll scale inconsistency. If you lack governance, you scale disorder.”

Scaling requires focusing on institutional strengths. Public institutions should focus on context-rich use cases rather than trying to compete with large technology companies.

Looking ahead, how do you expect AI to change the practice of justice?

AI will significantly increase productivity and enable the solution of more complex tasks. This will raise expectations for both citizens and public servants. We’ll have the chance to be more humane. At the same time, AI will shape the broader social context. It may create labour market tensions and increase demand for legal services. There are also risks from misuse, including fraud and cybercrime, as criminals adopt the same technologies.

What should public servants start doing now to prepare for that shift?

They should think about AI in strategic terms. Use it to solve meaningful problems, not just experiment with tools. AI should support work, not replace judgement. Used well, it can create space for more meaningful and humane work.

What excites you most, and what concerns you?

What excites me is the opportunity to reduce delays, reduce repetitive work and improve how systems connect. This creates time to be faster and more humane. What worries me is supervision. It’s difficult to oversee systems that may become more capable than their users. We need a better understanding and more interpretability to ensure we remain in control.