When the Universal Postal Union (UPU) unveiled its first-ever artificial intelligence (AI) agent at the 28th Universal Postal Congress in Dubai, it marked more than a technological milestone – it signaled the beginning of a data-driven transformation across the global postal network.

Developed with seed funding from the Japanese government, the pilot AI agent operates on the UPU’s Unified Data Platform (UDP), which has been designed to bring together vast quantities of postal and non-postal data. The AI agent’s purpose is to analyze, diagnose, and advise on postal development at country level, providing insights to strengthen postal reach, reliability and resilience.

As José Anson, the UPU’s Lead Economist, explains, this project represents “just the beginning of our journey in agentic AI and in data sharing.” He also notes that the UDP – developed to be fully integrated in UPU’s tech environment – is “a game-changer, because no such platform has ever been built in the postal sector before.”

Good data, better AI

The UDP is being constantly updated and improved, with a new prototype version due out at the end of November 2025. The aim is to ensure that it includes as much high-quality data as possible to power new agentic AI solutions, which will be built on top of the platform. “If you have garbage in, you have garbage out with AI,” Anson says. “There is no miracle. You don’t have good AI if you don’t have good data behind.”

To ensure quality, the UDP consolidates the UPU’s internal big data with other relevant sources – spanning trade, logistics and aviation – creating what Anson calls “a customized way of sharing data.”

“Data sharing has always been very difficult due to commercially sensitive data, personal data protection and other constraints,” he explains. “So, this platform is introducing a very granular way of data-sharing governance. We can embed each data-sharing contract inside the platform itself in an environment directly governed by the UPU.”

A collective intelligence

The AI multi-agent prototype showcased in Dubai offers a glimpse of how this data can come alive. In the future, it will form part of an ecosystem of specialized agents designed to “talk” to each other, each focused on a different dimension of postal policy and operations.

“One example of an agentic AI tool is specialized in the UPU’s Integrated Index for Postal Development (2IPD),” Anson explains. “This is the one we showcased at Congress. It knows everything about postal development diagnostics around the world.”

Another agent, he continues, is “specialized in Congress proposals and business plans – it knows everything about our plan for the next four years.” A third, the “consensus agent,” helps “find the right balance between different solutions or proposals,” Anson adds.

By connecting these agents, the platform allows the UPU – and, eventually, postal operators themselves – to model how policies and initiatives might improve real-world performance. In practice, that means a postal administration can use the system to identify its weaknesses and see how UPU programs can help close those gaps. “This is better than gen-AI,” Anson stresses. “Agentic AI acts, reasons and collaborates: it has real agency.”

Speaking every language

Perhaps the most striking aspect of the Congress demonstration was accessibility. “The experience was absolutely amazing,” Anson recalls. “For the first time, participants could understand the 2IPD diagnosis in their own language. This is a true game-changer.”

By supporting multiple languages, the AI agent ensures that data-driven insights are not confined to a few experts or technologists but can reach policymakers and postal executives around the world. “People could see, read and engage with complex postal data in their own language,” says Anson. “That changes everything.”

From prototype to global platform

For now, the agentic AI project remains in prototype stage. “We’ve sent it back to the lab for fine-tuning, and the next release will be on November 23 when the updated UDP is released.”

The overwhelming response at Congress has already created momentum for expansion. “Everyone wanted to know when they could start using it,” Anson says. “Ideally, we’d like to roll it out to all member countries early next year. But to do that, after the initial seed funding provided by the Japan Fund, we need to secure additional voluntary funding from governments and partners willing to sponsor the system’s expansion.”

The UPU is exploring several models, from donor sponsorship to subscription-based access. “One possibility is a freemium approach, where some parts of the service free, and others are paid,” Anson explains. “But we don’t want to become a software-as-a-service company. We’re not an IT provider.”

Instead, he envisions the UPU as a steward of a secure, ethical, collective intelligence ecosystem. “Agentic AI will actually facilitate private sector solutions, as well,” he says. “We want to offer a channel for Consultative Committee members and other partners to distribute their specialized AI agents through our platform.”

A secure foundation for collaboration

The Unified Data Platform and AI agents operate within the UPU’s trusted .POST top-level domain, ensuring a secure digital environment. “We offer maximum granularity of rights and permissions,” says Anson. “Countries have full control over how their data is shared.”

This governance-first approach distinguishes the UPU from commercial data platforms. “We’re neutral,” Anson emphasizes. “We’re not here to compete with the private sector. We want to facilitate collaboration, allowing everyone – from postal operators to private innovators – to benefit from the same trusted infrastructure.”

That infrastructure, he adds, is built on partnerships. “We’ve been working with Amazon Web Services on part of the data platform development using a cloud agnostic approach, with one of the most promising agentic AI startups in the world (Horizons Architecture), and with our own Postal Technology Centre, of course,” he says. “We’re a team of more than 20 people building this, combining internal expertise with external innovation and development.”

A layered vision

Behind the scenes, Anson and his team are looking at developing a sophisticated hierarchy of AI agents on the UDP. “There are three levels,” he says. “The macro agents specialize in designing better postal policies, regulatory frameworks and business models – the strategic level.”

Then come the “meso” agents, which focus on interoperability between networks, bridging systems and standards across countries. Finally, the “micro” agents handle practical solutions like route optimization or customs clearance allowing private sector providers to transform their current solutions into agent equivalents.

“What’s innovative,” Anson says, “is that we connect them all. The macro agents talk to the micro agents via the meso agents. This means we can align strategic vision with execution. That could be a tremendous game-changer.”

From data to development

Ultimately, Anson sees the UPU’s agentic AI initiative not as a tech project, but as a mission. “Our goal is to make the postal network smarter, more relevant and more resilient,” he says. “We’re using AI not just to analyze data, but to help countries make informed decisions about their postal future.”

If successful, the Unified Data Platform and agentic AI tools could become the backbone of a new kind of postal intelligence – one built on hyper-collaboration, transparency and shared innovation. “The postal sector has an incredible wealth of data,” says Anson. “Now, for the first time, we have the means to turn that data into collective intelligence – for the benefit of everyone.”

This article first appeared in Union Postale Autumn 2025.