Retailers are increasingly looking at how AI can help run their stores, but the big question is how much control to give over to AI and where human oversight is still needed. It's a topic that's been on a lot of minds at recent industry events.
The Shift Towards AI Adoption
Just a couple of years ago, most retail executives were hesitant to let AI take the reins, preferring to keep a firm grip on all operations. On a scale of one to five, where one is full human control and five is AI running everything, retailers were leaning towards a one.
Today, that number has moved closer to three and a half. This shows a significant and rapid change in how retailers view AI's role.
However, even with this shift, when presented with the idea of AI handling absolutely everything, most executives felt they weren't quite ready for that level of automation, or that it might not be the best fit for every single task.
Different AI for Different Needs
Companies like Quorso use various types of AI to help with store operations. For instance, one common use is monitoring operational data to flag when something goes wrong.
Consider a product recall. In such a situation, retailers want to be confident the product is removed from shelves quickly. For this, a more predictable, rules-based AI approach is preferred.
This means using AI that follows clear instructions and logic.
On the other hand, there are areas where retailers are more comfortable letting AI, like large language models (LLMs), optimize processes. For example, an LLM could be used to go through standard operating procedures (SOPs) and create a personalized plan to fix a specific issue.
This is where AI can really shine by processing large amounts of information and providing tailored solutions.
A Nuanced Approach is Key
Ultimately, the approach to using AI in retail needs to be nuanced. It's not a one-size-fits-all situation.
- Deterministic Models: Best for critical tasks where certainty is paramount, like product recalls. These are often rule-based systems.
- Machine Learning (ML): Useful for a wide range of optimization tasks where patterns and predictions are helpful.
- Generative AI (Gen AI) / LLMs: Excellent for tasks involving understanding and generating text, like analyzing SOPs or creating personalized action plans.
The goal is to use all these different AI tools together, integrating them into a single system. Being open about how AI is being used and what it's doing is also important for building trust and ensuring smooth operations.
