Coles eyes AI to keep shelves stocked in next viral recipe trend

Coles Group is hoping AI can play a role in better preparing the retailer and its value chain for the next viral recipe trend, after supermarket shelves were emptied of cottage cheese earlier this year.

Trending recipes on platforms like TikTok caused a run on cottage cheese and left supermarkets and their suppliers scrambling to meet demand.

General manager of data and intelligence Caroline O’Brien told the SAP NOW AI Tour event in Melbourne that Coles is moving beyond “really narrow, specific use cases” for AI “in siloed parts of the business”.

It will instead “prioritise” cross-functional use cases of the technology over the next 12 months, with viral product demand one of the problems that Coles wants to solve.

“There was a national cottage cheese shortage recently - it was some TikTok recipe trend - so in that world, how do we capitalise on social listening and sensing and then bring that into our demand forecasting, informing our supply chain, making sure that we’ve got the right promotions in store? How do we position ourselves into all of that,” O’Brien said.

“The idea [is] being able to react and sequence each part of our value chain to drive that opportunity, but also really practically manage that from a customer experience point of view, because if everyone wants to go in and get some cottage cheese because TikTok told them that it’s great, and they go into a Coles store and we don’t have the product on shelf, that’s [a] really core customer experience [problem].

“When we think about the supply chain, [we want to] have a supply chain that’s reactive to the demands we’re seeing driven.”

O’Brien - who is six months into her role as data and AI leader at Coles - said the retailer is taking a “considered” approach to its use of AI, favouring use cases that enable it to derive a competitive advantage in the market.

The retailer has grown its data and intelligence team to around 300 staff, supported by vendors including Microsoft, Snowflake, Databricks, Palantir and SAP.

O’Brien said the make-up of the team is becoming more software engineering-focused, matching the skills required to work with evolving artificial intelligence technologies.

“In building these technologies now versus eight to nine years ago, the profile of the builder is very different,” she said.

“What was very data science-heavy previously … is now a lot more software engineering-based. 

“So for us, we’re thinking about how to get software engineers who have experience with deploying a model, rather than having a data scientist who might have a little bit of software engineering experience. 

“We just think they can be a lot more flexible and adaptable in this environment and can solve 70 percent of the problems. They can get most of the way there.”