The Hype About AI Training is Massively Overdone for Retailers
Martec has recently done an exercise on the AI related training needs for each of the job roles in a retail business. Our conclusion is that the scale of budget some retailers are committing to AI training, or at least talking about, is way too high.
At headquarters, the IT Department may need more significant training, especially if they participate in building AI Agents, but not on the scale that many people believe and this is the reason. Most retail and consumer goods companies do not build the bulk of their own application software anymore. They buy it or licence it from companies like SAP, Oracle, Aptos, Microsoft Dynamics, RELEX, Robling and many more. These companies are all busy building AI Agents into their products. It is highly likely that they will be easier and cheaper to implement than using your IT Department with better functionality and less exposure to hallucinations that humans need to look out for. This is because they spread their investment over multiple clients.
To give an example, I read one supplier’s literature last week and they have so far built 14 AI Agents into their product set. One is to help a user make a better estimate of demand for a new product, which has no sales history. This is very common in fashion but applies on a smaller scale in many retailers. Historically, the buyer would think back to previous products in their assortments to determine which one would make the best proxy to use as the basis for forecasting demand. More advanced companies would tag every new product they introduced with attribute codes, so the selection of suitable products could be made a little smarter.
The AI Agent concerned takes this a step faster. It looks at all the available sources of product information in your company, such as web site descriptions of products sold, data in a product lifecycle system if you have one, customer database platforms, etc., assesses the input it gathers and then makes the best judgment on all the sources it found as to which ones would give the best starting point for the demand forecast. It would make little sense for a retailer that had access to this software and decided to build their own agent to do a similar process.
This is one example of many that I have read about in recent months.
Your IT Department will need to understand how to test and evaluate AI Agents for value, how to identify issues that need addressing and, in some cases, identify how to address them. So, they will need deeper training. But this is not something you will want to train lots of retail managers in or let them make any required changes.
There are two important conclusions from this. The first is that unless most of your core software is home grown, the challenge of developing and implementing an AI strategy may be a lot easier than you think.
The second is that many retailers (and others) are struggling to show any ROI on their AI investments. There will be a significant ROI from what your good suppliers are building in many cases. But to measure it and be sure, you will need a process for tracking implementation in the right way and a comprehensive, retail specific ROI model to do the analysis. Training built round extracting the value from your software investments will add to your ROI.
If you would like to know more about the process and tools to use, contact me for more information.-
Posted by Brian Hume
9th September 2026
