Why Agentic AI has yet to scale in retail

Walking the floor at NRF this year, one thing became clear very quickly.

While technology and services companies continue to talk enthusiastically about AI, many retailers are becoming tired of the noise. Not because they doubt its potential, but because too much of the conversation feels disconnected from day to day reality.

Almost every platform claims to be AI powered. Every demo promises transformation. Yet despite years of experimentation, only a small proportion of AI initiatives in retail have successfully scaled beyond pilots.

Retailers are no longer asking whether AI matters. They know it does. What they are questioning is why, after significant investment, AI has yet to scale in a way that fundamentally changes how their business operates.

The answer is increasingly clear. Agentic AI does not fail because the models are weak. It fails because the foundations are missing.

Most AI initiatives today are still deployed as isolated use cases. A forecasting model in one function. A recommendation engine in another. An automation workflow that never connects to execution elsewhere in the business.

These initiatives can deliver localised value, but they rarely scale across the organisation. As a result, AI remains fragmented, difficult to govern, and unable to operate end to end. Agentic AI is different.

It is designed to operate continuously, make decisions, and take action across multiple systems and processes. To do that, it requires a connected foundation.

A clear example is the Xfuze Sales Performance and Automatic Replenishment Agent. The agent continuously monitors sales, inventory, demand signals, and operational constraints across the business in real time. It identifies performance issues and recommends corrective actions based on shared, trusted data.

When supported by connected systems and processes, the agent can also trigger replenishment decisions automatically. By operating across systems rather than within a single application, it improves speed, accuracy, and consistency of execution, and can scale across the organisation to increase the return on existing AI investments. Importantly, many retailers are already part of the way there.

They have developed and launched AI agents. They have proven value in specific areas. What is often missing is not new agents, but the foundation that allows those agents to operate across the wider business.

Providing that foundation is not as hard or as disruptive as many expect. By connecting data, systems, and processes, retailers can unlock significantly more value from the agents they already have.

The same principle applies to the future of AI driven purchasing by consumers.

A related example is the growing interest in conversational commerce, where consumers can browse and purchase products directly within AI applications such as ChatGPT.

Retailers like JD Sports have already signalled their intent to enable this type of experience.

While the interface may be new, the underlying requirement is not.

For consumers to purchase seamlessly inside an AI app, product data, pricing, availability, promotions, and checkout must be accurate and connected in real time. Without a strong foundation, these opportunities stall quickly. The result is failed transactions, inconsistent experiences, and lost demand at the point of intent.

From conversations at NRF, retailers increasingly recognise a clear set of success factors required for Agentic AI to scale:

• A single, trusted, accurate foundation of data shared across the business
• Connected systems that allow agents to see and act end to end
• Connected processes and workflows so decisions turn into execution
• An orchestration layer to coordinate agents, manage dependencies, and govern outcomes

What is often underestimated is the cost of not putting these foundations in place.

Without a connected foundation, Agentic AI becomes a constraint rather than an advantage. Agents wait on fragmented data, disconnected systems, and manual handoffs, slowing execution at exactly the moments when speed matters most. Opportunities are missed, competitors move faster, and returns on AI investment are quietly diluted.

This is where platforms such as Xfuze play a critical role.

As a composable integration, data management, data orchestration, and analytics platform, Xfuze provides the connectivity across systems, data, and processes that Agentic AI depends on. It creates the shared intelligence and fabric that allows agents to operate across functions and, increasingly, across organisations.

The conclusion retailers are reaching is simple. If the business itself is not connected, Agentic AI’s cannot run and scale.

Retailers are not tired of AI. They are tired of AI that never moves beyond experimentation. The next wave of value will belong to those who pair their Agentic AI investments with a connected foundation that allows agents to operate across the business, rather than in isolation