The Invisible Infrastructure of Commerce: OpenText's John Radko on Business Networks and the Rise of AI Agents

Every day, companies exchange an enormous flow of information with their partners. Purchase orders, sales orders, shipping notices, invoices, and countless other documents move between organizations in a constant stream. It is, as OpenText's John Radko puts it, the lifeblood of commerce. And most of the time, it happens invisibly.

In a recent conversation with Constellation analyst Chirag Mehta, John Radko, Senior Vice President of Product Development at OpenText, explained what makes that flow possible, why it is so much harder than most people assume, and how AI is beginning to change the picture.

The problem business networks solve

The core challenge is simple to state and difficult to solve: companies need to talk to each other, but no two companies are alike. They run different systems, different generations of technology, and different data formats. Even when two organizations use the same document standard, they often apply it differently, including omitting values or relying on entirely different versions.

The result is that a single process, something as routine as placing an order, can look ten different ways across ten different partners. A business network exists to mediate all of that complexity, connecting organizations and keeping information moving reliably between them.

And the complexity does not hold still. As Radko notes, the moment a company changes its ERP system or adjusts how it handles receipt of goods, that change can cascade across its entire partner ecosystem. The network has to absorb that constant motion.

Onboarding and the build-versus-buy question

Getting partners connected is one of the central capabilities OpenText offers. Smaller companies without communication gateways might connect via a web interface, while large enterprises connect directly to SAP or Oracle. Increasingly, mid-market companies are connecting via cloud ERP platforms such as NetSuite and Microsoft Dynamics.

So why not just build all of this in-house? Radko is refreshingly candid here. There is very little, he says, that clients could not do for themselves, just as they could run their own logistics or cater their own events. The real question is whether it makes business sense. For most organizations, B2B integration is not anyone's full-time job. For OpenText's business network team, it is the only job. That focus, combined with purpose-built technology like the Trading Grid platform and years of refined processes, is where the economies of scale come from.

Managing at scale

When customers are processing hundreds of thousands of data exchanges a day, things inevitably go wrong. A partner goes offline, then comes back and tries to retry thousands of pending transactions at once. Handling that without overwhelming the system requires more than connectivity and data transformation.

That is why a large share of OpenText's software investment goes into what Radko calls community management: software that monitors transaction flow, lets customers manage their partners, and surfaces problems quickly. Their TG Insights product, for example, lets customers search through millions of documents to find a specific transaction, while automatically alerting them to failures across their system.

From acceleration to transformation with AI

On AI, Radko frames the journey in two stages: it begins with acceleration and ends with transformation. Today, AI is speeding up nearly everything OpenText does, from onboarding partners and translating data to resolving issues and detecting changes across the network.

The bigger shift is still ahead. OpenText is deploying AI agents that help clients manage their communities, onboard partners faster, and resolve issues more quickly. Looking forward, Radko sees agents acting as managers of the community and a genuine first line of defense. He points to a recent capability rolled out with a major automaker, where the system detects a file failure and automatically resends the file once the underlying issue is resolved. Simple on the surface, but it requires real knowledge of what is happening across the network to know when the retry will actually succeed.

Radko believes the majority of e-commerce and B2B activity will eventually be performed by agents, a view increasingly shared across the software industry as the conversation shifts from end users to agents.

One piece of advice

Asked what he would tell organizations to do right now, Radko's answer is direct: treat partner connectivity as a platform. Consolidate as much as possible onto a single infrastructure, because every partner you connect to the network unlocks all the capabilities of that network. As agentic AI rolls out, it will work through the interfaces and APIs available across your services, so the more partners and processes you bring into one place, the more scale and momentum you can build.

In other words, the groundwork organizations lay today will determine how ready they are for an agent-driven future. The business networks that do the invisible work of commerce are about to get much more intelligent.