How AI is redefining retail promotions

Want the latest analysis?

Subscribe to our newsletter for the latest trends, research and insights in Retail

Most retailers are familiar with how AI can be used to power their marketing strategies. Most marketing systems now have personalised send schedules that work out when customers are most likely to respond to a promotion. The more obvious methods include working out when’s the best time to send a push notification, dispatching an email when there’s the greatest chance of it being opened by a customer, and what products to recommend within that email based on a customer’s profile. These tasks are done with algorithms, plus a sprinkle of AI to optimise them. However, there are a number of rapidly emerging less well-known AI-driven promotion techniques…

Smart in-store signage

Retailers are beginning to install in-store digital signage that can be used to promote their own products and those of other brands, to which they sell the ad space. More exciting is what’s happening with cameras powered by AI. They can detect not just gender, but also a customer’s apparent style or the kind of products they are wearing. They then change in-store sign displays based on what they detect with the aim of attracting the attention of each customer that walks past and inspiring a purchase.

This trend is likely to move relatively slowly because there's a privacy element to it around how much customers want their data to be used. But certainly basic personalisation and promotion of signage is going to happen.

Monitor the shop floor with computer vision

Retailers are also incorporating AI into CCTV cameras already in place across their stores. This is more cost effective than buying new kit and can be used to create heat maps of customer movements to help with store planning and merchandising. Perhaps a product's not selling well because it’s in a low footfall area? Or should it be dropped if it’s in a busy area but not selling?

Another application is detecting hazards in store. For example, monitoring aisles of a supermarket to detect spills or obstacles. There’s also a stock dimension, using smart cameras to spot empty shelves and notify staff to restock.

Boost merchandising effectiveness

Web and app merchandising is increasingly controlled by algorithms and AI including personalising sort orders such as ‘You May Also Like’ product similarity recommendations, personalised product feeds and AI-generated merchandising lists. And some of the highest return on investment is being seen from AI customer experience features like these.

You can also use computer vision to better understand a product. For example, if you've got poor metadata that only identifies the item and its colour, AI can work out, for example, the length, the style and more. It can also find other similar products, such as blue dresses in the same length that are similar, which is ideal for the recommendations mentioned earlier.

A lot of people come into a website through Google search and the item they find might not be in their size or include a feature they don't like. Presenting the customer with similar products can retain their attention and inspire an alternative purchase.

Create catwalk videos and colour variants

AI can create catwalk-style videos for products to a surprisingly high quality right now. This can have multiple benefits, including accelerating a product to market, which is particularly important in industries like fashion. It’s also cheaper than shooting real models on the runway, and far more engaging than static product shots.

Drawbacks include making AI models that are ‘too perfect’ as this can set harmful and unrealistic beauty standards. Plus, the quality can be a problem for particularly complex or expensive garments. The non-deterministic nature of AI can also be incredibly frustrating, and although software-as-a-service AI models are quite cheap now, they are likely to go up in price.

AI can also be used for retouching images, using a single product shot and changing the colour to match other variants. Again, this can be done surprisingly accurately, saving the time and cost of taking multiple product shots. However, care must be taken to match retouched images as closely as possible to actual product colours. Getting it wrong can damage customer satisfaction and confidence.

Find out more about retail tech trends in our upcoming Spring 2026 report.

Want the latest analysis?

Subscribe to our newsletter for the latest trends, research and insights in Retail