As one of Britain’s fastest growing and most innovative retailers, supermarket chain Iceland uses data and AI to enable “business at the speed of thought”.
To surface information to store colleagues, Iceland used Azure OpenAI to create Genie, an app which enables the knowledge base to be queried in a conversational way.
The Genie app is helping colleagues find answers faster. Its success is inspiring the Iceland team in further AI innovation with Microsoft Azure and Azure OpenAI.
Our use of Azure OpenAI has just begun, and we can already see numerous opportunities where we feel it will deliver business value, transforming the way we can interact with and train our colleagues.
Louise Dhaliwal, Chief Information Officer, Iceland
As one of Britain’s fastest growing and most innovative retailers and one of the best UK companies to work for in 2024, Iceland Foods employs more than 25,000 staff across its 950 UK locations. For many years Iceland has led the UK market in frozen food innovation and innovation is written in every strand of the company’s DNA, including in its approach to technology.
In order to realise the leadership team’s ambition to “do business at the speed of thought”, Iceland’s technology team is on a mission to improve the data architecture of the business and present insights in meaningful and actionable ways.
“We have a huge amount of knowledge in the business, but it is spread out all over the place – Word documents, personal file shares, inboxes and, particularly, people’s heads,” states Stuart Bickley, Head of Application Development at Iceland. “We wanted to capture and consolidate that knowledge.”
To achieve this, the Iceland team has built an ingenious new tool through which colleagues can interact with business knowledge – an app called Genie which leverages the natural language capabilities of Azure OpenAI.
Craig Robinson, Software Engineering Manager at Iceland, explains, “We’d had this idea of consolidating our knowledge base. Then when Azure OpenAI came along, we realised it offered us a way to enable users to interact with that knowledge using natural language.”
“Azure OpenAI seemed like a natural fit for us,” agrees Stuart Bickley. “It would fit easily within our existing app landscape and, as a platform as a service, it is easy to set up and configure – and it’s all there for us wrapped up in the Microsoft infrastructure and security ecosystem.”
Iceland has a tradition of creating key line of business apps in-house, opting to build on the Microsoft Azure technology stack.
Craig Robinson explains, “In Azure, Microsoft is offering the best technology that is available now. It’s all highly available. It’s very configurable. And I really like Microsoft’s well-architected framework, which helps us make the right decisions around resiliency and cost.”
This in-house app development programme began more than a decade ago with the development of Iceland’s Nexus HR app and now spans financial, product specification, scheduling apps and more. Genie is the latest innovation.
“We’re really hands on at Iceland and we get to work on some really interesting projects,” explains Craig Robinson. “For Genie, we built our own drag-and-drop content builder so the retail team could create the source documents in a great-looking format. Genie runs in Azure Container Apps and uses a Cosmos DB no-SQL database. We use Azure Cognitive Search for indexing and Python functions in Azure to interact with Azure OpenAI. Iceland’s culture of innovation gives us that opportunity to use the latest robust, enterprise tech that’s new out-of-the-box from Microsoft.”
“At first there was a lot of manual testing and iteration, working closely with the business,” explains Stuart Bickley. “It was a fine balance to try to make sure Genie wasn’t too formal and was providing completely accurate information for our employees.”
Genie was given its first test in December 2023. Iceland started small: making its 2023 Christmas guides for store teams searchable through the Genie app installed on the instore PCs of Iceland store managers.
The retail operations team at the head office created the pages which would form the basis of Genie’s knowledge base. Lee Boswell, Consumables and Process Manager at Iceland, explains, “Our annual Christmas guides tell store managers about staff planning, stock control, trading hours – all the things they need to run the store during the busiest time in retail.”
Previously, this information had sat on a SharePoint knowledgebase or was printed and pinned to the walls instore. Now, staff can ask natural language questions of Genie to understand what they should do.
“As soon as the team saw what Genie could do, they were really excited,” says Louise Dhaliwal, Chief Information Officer at Iceland.
Louise Dhaliwal states, “Our use of Azure OpenAI absolutely has got legs. It’s made a huge difference to how we can interact with and train our instore colleagues.”
Not only can searches use natural language, rather than being limited to exact terms or fuzzy matches, answers are more targeted and more concise – making the experience faster and more streamlined.
“Genie presents you with the summarised answer and, beneath that, provides the link to the source documentation,” explains Lee Boswell. “It’s so much faster than a traditional search. Instead of being presented with links to 15 different PDFs and then having to scan them all to see if one contains the answer you need, you are presented with the answer immediately.”
“It is really impressive the way we can interact with Genie,” confirms Stuart Bickley. “Those early results drove us down the path to make improvements and extend its scope.”
“Once we’d trained people to use Genie for their Christmas planning queries, we wanted it to be a fast follow-on,” says Louise Dhaliwal.
Eager to build on the momentum of success, the Iceland retail operations team quickly added new and additional content, including training guides, policies and process documentation.
The project has given many different colleagues an opportunity to develop their skills. Lee Boswell enthuses, “I started as a store manager at 18 and later moved into the head office retail operations team. This project has been a real eyeopener and I’ve loved getting to learn something new.”
“The fact Genie was giving the right answers and taking us to the right places was a lightbulb moment for me,” continues Lee Boswell. “It encouraged me to learn more about how I could build the pages so that, if colleagues clicked through, the source material was really engaging. We’ve worked very closely with our colleagues in IT to keep improving and developing the solution and the way information is presented.”
“We’ve now rolled out Genie so it is being used throughout the business and people love it,” states Stuart Bickley.
“We’ve started this journey early, but the teams have embraced it,” agrees Louise Dhaliwal. “We’ve still got a long way to go, but we’re learning every day. Genie has been a very exciting project. There’s a whole raft of business functions we can move down the same route, including across our HR and payroll systems.”
Lee Boswell agrees, “It’s so much better than what we had previously and we’re only just scratching the surface of what’s possible.”
“We’re looking to plug Genie into dynamic data sources, like our Nexus HR system,” Stuart Bickley elucidates. “People will be able to ask ‘how many holidays do I have left this year?’. Our CEO is very interested in how it can be applied to business statistics, so we can ask ‘what were the top selling products last week?’ or ‘what were the sales in London?’ for example.”
Another AI innovation is helping Iceland’s software engineering squads to deliver these new capabilities for Genie, as well as for other apps. Iceland is using GitHub Copilot to boost development delivery velocity.
“A few of our developers, including myself, had a personal licence for GitHub Copilot so we became big advocates for it. When the enterprise licence became available, we snapped it up!” reports Craig Robinson.
“GitHub Copilot has been fantastic for supporting the developers who are more familiar with C# to work with the Python functions we’re using. It’s also given the junior developers something to lean on so they don’t need to ask their senior developers as many questions. We’re getting really good feedback on it.”
Craig Robinson continues, “The power of GitHub Copilot is that it can do so many different things. Copilot has also really helped out in getting more unit tests into our apps – there’s no excuse not to include them now. That’s why we’ve given every one of our developers access to GitHub Copilot.”
Our use of Azure OpenAI absolutely has got legs. It’s made a huge difference to how we can interact with and train our instore colleagues.
Louise Dhaliwal, Chief Information Officer, Iceland
The IT team has leant into the learning content provided by Microsoft. Craig Robinson explains, “Some of my team have completed the Microsoft Certifications, which has been great for them and their personal development. And, obviously, it gives them the skills they need to delve into this exciting technology.”
The close partnership with Microsoft has also helped the Iceland team feel confident in their innovation. Craig Robinson says, “Microsoft has given us enough support so that it’s not overwhelming. We know if we get stuck, there is somebody we can talk to, who can unblock us and point us in the right direction.”
He continues, “Microsoft has been great to chat to and bounce ideas off and hear what is coming up as well. We’ve leant into the support available to us through our Microsoft Unified Support contract and Microsoft has helped us to follow best practice, right from the start. That’s been important, especially when setting up Microsoft Fabric.”
As part of Iceland’s strategy to enhance the data infrastructure of the business and enable rapid, data-driven decision making, the team has begun implementing Microsoft Fabric, the AI-powered analytics platform that unites data and services.
“We want to consolidate all our business data under one roof,” explains Craig Robinson. “We’re going to utilise everything that’s within Microsoft Fabric to build custom semantic models over the data and to get the maximum amount of value from our data.”
“You can’t just plug in AI. Your data has to be in a ready state,” emphasises Louise Dhaliwal. “Having that consistency, accuracy, completeness and trust in the data is vital – and that is where Microsoft Fabric will deliver for us.”
At the same time, the Iceland team is exploring further AI innovation in a new business app which will draw on the data in Fabric. The new volume forecasting app will use business data to predict sales patterns so store managers can better schedule staff working hours – and stores are not over- or under-staffed.
“Instead of relying on store managers’ intuition, we are using the AutoML in the Azure AI Studio to create a machine learning model which runs time series analysis by hour,” explains Craig Robinson. “Ultimately, it will optimise the way stores operate by improving our forecasting accuracy.”
Louise Dhaliwal states, “We’re quite a lean team at Iceland so we rely on our partners to become part of our team. Our engagement with Microsoft has been really good.”
Of course, the work hasn’t been without its challenges. Craig Robinson says the biggest challenge for his squads has been adapting to the difference in the development lifecycle when working with generative AI.
“When working with LLMs, you need a mindset shift from everybody – the developers, the business analysts and the test team,” he posits. “It was a particular challenge to test the LLM due to its varying probabilistic behaviour when compared with traditional software testing. And that understanding requires a big cultural change.”
Stuart Bickley agrees that the close relationship with Microsoft and the collaboration between IT, the business and Microsoft has been pivotal. He says, “We have a great relationship with Microsoft and its really, really useful to have them as a sounding board to make sure we’re doing things the right way, that we’re following best practice and we’re not going down any dead ends – because this technology is moving so quickly.”
“The advancements we’re seeing in Azure OpenAI in particular, but across all areas of machine learning, and the offering we’re getting from Microsoft are moving so quickly,” continues Stuart Bickley. “We can already see the other business problems where Azure OpenAI could be applied. We believe it can help us to achieve things much faster than we’ve been able to achieve before.”
The breadth of work underway at Iceland is inspirational and exciting for the teams involved. It is made possible, in part, by the culture and dynamism that characterise Iceland.
“We don’t have an innovation team at Iceland,” states Louise Dhaliwal. “We have innovation inside all our teams. We’re not afraid to try and we have a very innovative board.” she also states that, “Our use of Azure OpenAI has just begun, and we can already see numerous opportunities where we feel it will deliver business value, transforming the way we can interact with and train our colleagues.”
“Our food may be frozen,” smiles Stuart Bickley in agreement. “But our tech stack is never frozen.”
Follow Microsoft