The AI Mirage: Are You Investing in Real ROI or Just Expensive Reports?
The AI Mirage: Are You Investing in Real ROI or Just Expensive Reports?
In the retail profession, we’re all currently navigating the hype around AI. We are sold sophisticated tools, which cost a lot to run and sometimes produce incorrect results. Sometimes we can perform the same tasks equally well or occasionally better without using AI. Retailers are learning that a lot of current AI applications are saving staff some time, but not having any impact on company results. This mindset, while common, is partially flawed. And this uncritical acceptance is leading to significant, wasteful expenditure.
The current hype around Generative AI, in particular, is creating a mirage. It is often a source of expensive reports and unfulfilled promises. When your people can't tell a useful tool from a gimmick, the business ends up paying for a very expensive learning curve.
Agentic AI is the latest hot topic and most people will come into contact with them as chatbots. One day these "agents" may be very helpful. Today they are mostly very frustrating to use.
Finding the Real ROI
The first thing to remember is that 'AI' is not one single thing. We must differentiate between headline-grabbing GenAI and the powerful, practical machine learning tools that are already proven. When your teams have the right skills, they can leverage these tools to protect your P&L in three critical ways:
1. Improving Forecasts and Inventory This is where machine learning is already a proven, powerful tool. A good algorithm can analyse vast sets of data—promotional calendars, historical sales, even weather patterns—to create a demand forecast far more accurate than a human-run spreadsheet. If Merchandising teams ignore these tools, or worse, do not know how to interpret their output, their replenishment models will continue to be flawed. This leads directly to costly overstocks or missed opportunities.
2. Reducing Fraud Computer vision is a branch of AI that is working today to reduce fraud or theft at self-checkout stations. They spot situations where someone declares an apple as a type with a lower price or they have left an item in the bottom of their cart to avoid paying for it. To quite a high degree, though not yet completely perfect, facial recognition can identify past shoplifters coming back into a store to piotentially steal again.
3. Spotting the 'GenAI Mirage' We now see vendors selling GenAI to 'automate' buying or write thousands of product descriptions. The problem is that these tools often lack deep commercial context. A GenAI does not understand your 'Good, Better, Best' strategy or a vendor's true cost structure. It can write a product description, but it cannot know why a customer needs to be convinced of the 'Better' product's specific value. This is a direct path to poor sell-through and future markdowns.
4. Minding the 'Human-in-the-Loop' Skill Gap The biggest mistake is believing AI replaces your team. It is a co-pilot, and a co-pilot is useless if the pilot cannot fly. Unfortunately, many vendors are selling these tools without the necessary industry training. Your team ends up with a complex dashboard they do not understand. They cannot spot an 'AI hallucination'—like a nonsensical forecast—because they lack the deep, foundational retail skills to know it is wrong. The AI's bad data is then treated as fact, leading to expensive buying and inventory errors.
Conclusion Stop treating AI as a magic wand. The real value today is not in the most expensive, headline-grabbing application. The real, bankable ROI is in using proven machine learning tools to solve specific commercial problems, like forecasting and replenishment. This requires a structured approach, and it starts with equipping your teams with the advanced data literacy and, most importantly, the foundational retail skills to understand what the tools are really telling them.
If you would like to explore a detailed framework on how to use practical data tools to improve your forecasting and buying cycle, you can find it in our Retail and Consumer Goods Industry WIKI.
-
Posted by Martin Dugan
26th November 2025
