Why modern ERP is critical to scaling AI agents
AI agents are only as good as the processes and data beneath them.
In retail, AI agents are increasingly used to carry out a wide range of tasks. There are agents that forecast demand, support store staff or adjust prices. At the same time, consumers are turning to AI agents to find and compare products. A shopper might ask an agent, for example, to find a specific product that can be delivered before a certain date. The agent then needs fast and reliable answers on price, stock and delivery time.
Yet while this is quickly becoming common practice, many retailers struggle to move beyond isolated AI use cases. The challenge is rarely the agents themselves. More often, they expose underlying issues in processes and data that have existed for years.
And for both, a modern ERP provides a critical part of the operational foundation, connecting the processes and data that agents depend on.
Why so many agents fail to deliver
At many retailers, the data that agents rely on is fragmented. Stock, prices, orders and product information are spread across systems, each with its own definitions. The data is often good enough for daily operations, but not consistent enough for an agent to rely on.
The root cause often lies in the ERP landscape. Many retail ERP environments have grown over decades. Some companies run several versions of the same ERP system, even when their business model does not justify that complexity. Years of customization and manual workarounds mean that stores, e-commerce and wholesale work in their own way. And because processes create data, every deviation in the process becomes a deviation in the data.
“We see many retailers that are eager to start working with AI agents, but haven’t put the foundation in place,” says Ernst Stoelhorst, Managing Director Retail at IG&H. “The result is a scattergun approach: plenty of initiatives, but very few that hit the mark. And when the data underneath is fragmented, you cannot even tell whether what an agent tells you is right.”
At the same time, the pressure is rising. Vendors are phasing out many widely used legacy ERP systems, and a full redesign can take two years or more. Meanwhile, retailers that already run on a modern core can deploy their own agents faster and give their customers’ agents more reliable answers. And as agents become more common, that advantage only becomes more important.
That makes ERP modernization more than a technical question. It is increasingly a question of competitiveness. Modernizing the ERP touches processes, ownership and the way the organization makes decisions. That is why the decision cannot remain within IT. It belongs at board level.
For that board-level discussion, it helps to be precise about what AI agents actually need. They place higher demands on processes and data than traditional systems ever did. In 2025, Gartner predicted that through 2026, organizations would abandon 60 percent of AI projects that are not supported by AI-ready data [1].
For retailers, four requirements make the difference
Consistent processes to act within
Agents do not only read data. Increasingly, they also act: placing a replenishment order, adjusting a price or handling a return. That only works when the process behind it is the same everywhere, with clear rules, approvals and exceptions. An agent cannot follow a workaround that only exists in someone’s head.
Real-time instead of batch
In many legacy landscapes, stock, orders and prices are synchronized between systems in batches, sometimes only once a day. For a report, that is fine. For an agent that has to tell a shopper whether a product can be delivered before a certain date, it is not.
Trusted and consistent data
When point of sale, warehouse, planning, finance and promotions each maintain their own version of the same data, an agent receives conflicting answers. Which stock figure is leading? Which price applies?
Clear definitions and ownership
An agent cannot guess what “available” means or which system is right. Someone has to decide which source is leading, which definition applies and who is responsible for keeping it correct.
None of these requirements are new. In fact, they are the same challenges ERP modernization programs have been addressing for years. The difference is that AI agents make it impossible to ignore them. Retailers can no longer rely on workarounds, fragmented processes or inconsistent data when decisions and actions increasingly happen at machine speed.
The role of ERP in building the foundation
A modern ERP is where these requirements come together, and it keeps them in place as technology evolves. It is where core processes run, where much of the data is created and through which agents increasingly act.
Processes run in the ERP
When every channel follows the same standardized process in one core, agents have clear rules to act within. Consistency is also built in at the moment data is created, which is far more effective than repairing it afterwards.
A modern core works in real time
Modern ERP platforms process transactions in real time and offer standard interfaces. Together with systems such as order management and product information, they give agents direct access to reliable data. That opens the door to use cases such as demand forecasting, pricing and replenishment.
The transformation itself is the moment to clean up
Moving to a modern ERP means reviewing, cleaning and redefining master data anyway. Retailers that use this well build their agents on a clean foundation. Those that skip it simply move the mess to a new system.
A clean core keeps pace with innovation
In a clean core, the ERP itself stays as close to the standard as possible, and customizations are built around it rather than into it. That makes upgrades far less disruptive. Retailers can adopt new capabilities from their vendor, including new AI features, much sooner. Moreover, they can roll out their own improvements without breaking what already works.
Creating the foundation for AI at scale
Building this foundation requires more than implementing technology. It requires ERP, processes, data and AI to be approached as one transformation.
At IG&H, we support transformations to modern ERP landscapes and help organizations standardize the processes that run through them. We also help retailers improve master data, define ownership and governance, and design and build AI agents that fit the way their business operates.
We help retailers connect ERP modernization, process standardization, data readiness and AI in a single transformation. This creates a strong operational foundation for AI, allowing successful use cases to move beyond experimentation and scale across the business.
Ready for what comes next
Adding AI on top of a fragmented landscape will only get retailers so far. Agents rely on what sits underneath: connected systems, reliable data and well-designed processes. In an agent-driven market, that operational foundation could become the real competitive advantage.
Sources
Looking for help discovering your future landscape and building your transformation case?
It starts with a conversation. Get in touch
Ernst Stoelhorst
Managing Director Retail & Products
+31653440731
