AI In The Retail Sector | Peak Indicators

AI in retail hasn’t really been about customer-facing innovations because immediately accessible large value can be achieved in the capabilities AI is bringing to the backroom to help optimise operations and gain greater insights into shoppers’ behaviour. That’s phase one of AI in retail and many retailers have completed or are well on the way to adopting AI for various backroom systems — and they’re reaping the benefits. The next step is to become more shopper-facing and several different approaches are being taken. These are broadly split into the transformation of traditional retail via new approaches such as the autonomous store or the experiential applications of AI for things like augmented reality mirrors, interactive walls or tactile experiences that need to be measured. As always, and especially in these times of pandemic, retailers need to consider their strategies carefully and understand the pitfalls to watch out for, what constitutes best practice and how to manage expectations.

How advanced is AI in retail?

I think there’s probably a difference between key features and nice-to-have enhancements. I would say that for many core retail tasks, then the mature AI technologies are working well in many situations. The predominant use case is likely predictive analytics based on machine learning. This includes online and offline scenarios. Many companies are experimenting with AI techniques for more advanced user experience features, including recommendations. Companies, such as Amazon and Asos, are also using robotics etc. to automate their warehouses.

Is utilisation of AI in the backroom of stores starting to enable retailers to move AI onto the shopfloor?

I’d say it’s worth reflecting on what you mean by ‘the shopfloor’. As we emerge from the first wave of a major pandemic then it’s likely that the shopfloor is now virtual and online compared to physical. There are certainly many features provided on retail websites that are powered by AI (esp. ML), such as recommendations or digital assistants, that are now commonplace.

What AI applications do you see on the shopfloor (any examples would be helpful)?

In the fashion industry, companies like Asos and Zalando are starting to use advanced image recognition and mobile apps. You can take a picture of an object or someone you see on the street and the app will use visual analysis and machine learning to show matching items from a product catalogue or make recommendations of what products shoppers may also like.

Is AI enabling concepts such as autonomous stores? What else is needed — AI alone isn’t the complete solution, is it?

My view is that AI and related technologies should support (and make more efficient) existing human/business processes. In this sense the autonomous store is one in which people and machines work together. So no, AI is not the complete solution. This could involve educating consumers and cultural change in the sector/business. AI is a tool in the toolkit for developing more effective and efficient processes and enriching users’ retail experiences.

What are the challenges of measuring success of shopfloor AI?

There are likely many challenges in measuring success of AI that is customer facing (i.e. on the shopfloor). Understanding what success means is a starting point, as well as developing suitable criteria and metrics for success to monitor and track success. There will also be questions around ethics and privacy if user data is being used to drive metrics. There is also the challenge of integrating experimental methods into the entire process, e.g. tracking reliably success over time and using techniques, such as A/B testing, to see the effects of using particular AI techniques or introducing new features.

To what extent does AI in the backroom need to integrate with shopfloor AI?

These can likely be treated as separate, but in the future having integration between the backroom and shop floor would allow real-time and more responsive applications to be developed. This would also allow for inputs from the shop floor to be fed back into the backroom processes, such as real-time stock monitoring.

How can an ROI for AI in retail be constructed, especially given that many projects are experimental?

The ROI might not come directly from sales but indirectly, e.g. popularity on social media. Understanding the contributions and benefits of AI (along with the costs) will help define the right metrics for ROI. It might also be hard to determine whether the contributions of AI to sales and attention must be paid to cause and effect relationships.

What does best practice for AI in retail look like?

I think best practice is identifying the use cases where AI can be applied and then experimenting with technologies to see which are doable. Getting the foundations sorted first is also a key requirement as without this you are simply feeding poor quality, biased or inaccurate data into any AI system (which will simply output low quality results). Developing interest amongst the senior management of the business is also key so that buy-in is obtained early on for any AI initiative. Bringing in people with the right skills is also key to success.

Originally published at https://www.peakindicators.com on January 2, 2022.