From securing stakeholder buy-in to deciding where an AI team should reside within an organization, retail tech leaders from Europe, the Middle East, and Africa came together to explore how Data & AI are transforming the industry and to share strategies for overcoming common challenges.
AI is on top of every business’s agenda, particularly in retail where emerging technologies impact everything from CX to inventory to merchandising decisions.
The latest Logic Retail Leaders Forum brought together a dozen retail leaders from across Europe, the Middle East, and Africa to discuss how they are implementing data & AI solutions into their organizations and overcoming challenges.
The session was hosted by two seasoned retail tech leaders from Logic, Part of Accenture: Graeme McVie (Global Leader of the Data & Analytics Practice) and David Hawkings (Head of Logic Europe, Middle East & Africa) who have decades of collective experience successfully implementing complex AI/ML initiatives with a wide variety of retailers around the world.
The Forum participants had a lively and frank conversation about the current state of AI within their organizations, the decision-making processes they use to prioritize competing Data & AI/ML initiatives, their experiences implementing changes, and the next steps they plan to take. Here are six key takeaways from the discussion:
1. It’s important to separate AI hype from reality
AI is an umbrella term with many sub-branches including everything from expert systems to Artificial Narrow Intelligence (with generative AI as a further subset within ANI) to Artificial General Intelligence (AGI) which remains, for now, in the realms of science fiction.
Some ANI technologies, such as computer vision, are well-established and reliably add value (e.g., shelf inventory management, loss prevention, and consumer-facing AR applications), while others are experiencing their moment in the hype-cycle sun—or at the “peak of inflated expectations” as Gartner analysts talk about it.
It’s easy for non-technical users to be swayed by the new technologies they hear about in the media. The reality, however, is more complicated: Tech leaders must set expectations for the rest of the organization and communicate the viability of a new technology to deliver value in its current state of evolution.
Adding more confusion, there seems to be some willful conflation on behalf of AI vendors who mix terms like genAI and AGI. “I often end up in conversations with retailers where people think generative AI is AGI,” Graeme noted. “And I have to explain the difference to keep the conversation focused on realistic AI/ML capabilities that can truly add value right now.”
2. The importance of leadership buy-in
The Forum kicked off with a brief exploration of a pair of examples of European companies implementing AI analytics into their org: a consumer electronics retailer using GenAI to personalize product information suggestions and to enhance the post-sales customer experience, as well as a large grocery retailer that is developing an insight platform for customers and suppliers that will lead to personalized and tailored experiences.
This was a jumping-off point to highlight that not all companies (or IT leaders) enjoy the same organizational buy-in when it comes to Data & AI/ML. Forum participants remarked on a range of experiences related to their organizations’ commitment to—and understanding of—Data & AI/ML; the maturity of their Data & AI/ML strategy; and the deployment & use of Data & AI/ML capabilities across their organizations.
The AI journey can be illustrated in five stages:
Interestingly, no retailer in the Forum said they felt like they were at Stage 1, and while some felt like they were approaching Stage 4, not one felt like they had arrived at Stage 5 yet.
3. Everyone is on a Data & AI/ML journey, but frequently they’re on different paths
To move companies along their AI journey, IT teams must demonstrate the value of these new technologies and make them accessible to the entire organization. The forum participants used the framework of the following four-step Data & AI/ML journey to describe where they had started and where they would like to go next:
One participant said he was brought in to help “play catch up” with AI. “The company realized that we were somewhat late to the party when it comes to data and insights and this is why they hired me,” he explained. At that stage, their initial goal was to “build a robust foundation for the future of data.”
Implementing these technologies can be a far more involved process than many business leaders understand. He went on to describe how he balanced taking on all the necessary foundational data work while providing the management with “shiny objects,” i.e., tools that could deliver actionable insights to business users.
“The business doesn’t want to wait 12, 18, or 24 months for us to put in place an enterprise data warehouse and data governance,” another participant explained their experience with this common challenge. “They want insights immediately, so we have to work out how to build the plane while we are flying it.”
A different participant, an IT leader with an e-commerce retailer, described the importance of making the benefits of AI available to as much of the organization as soon as possible. The key, they explained, “is filtering all the data into really usable streams for other areas of the business to understand the insights… and make it accessible to ‘the people making the decisions.’”
Another LRLF participant highlighted the fact that you don’t have to wait for perfect data before being able to deliver business value: “I sometimes hear people saying that you can’t deploy AI/ML solutions until you have perfect data. That is not true—and besides, if you wait until you have perfect data within a retail environment then you’ll be waiting forever.”
A discussion point was around the prioritization of competing Data & AI/ML initiatives, which would be different for organizations that sell different types of products (e.g., fashion versus food). The participants found a common ground that it is important to have some decision-making framework in place to assess the value and ease of deployment of data/AI/ML capabilities, such as with the following framework:
4. Delivering high-value use cases is important for Data & AI/ML momentum
The forum participants were very engaged on the topic of identifying and deploying proven and valuable use cases. One of the most readily applicable and highest-impact use cases for AI/ML in retail? Inventory forecasting.
A participant mentioned their experience with a seasonal retail business that was having issues with misaligned inventory—either having too much or not enough. They went on to mention that not only can AI/ML be used to more accurately determine demand forecasts that would inform more accurate initial order quantities, but it can also guide inventory allocation to individual stores, implement markdown optimization capabilities, and intelligently manage excess inventory at the end of the season to achieve their sell-through, revenue, and gross profit targets.
Another participant described a forecasting application with a major European cosmetic brand, in which the tech not only predicted demand and guided replenishment for existing products, it incidentally provided insights for new products. “They were using lots of data feeds including trends,” they explained. “And the interesting thing is it highlighted that it didn’t have enough vegan makeup in their mix, which it wasn’t supposed to do. That wasn’t the exam question. But funny enough, they were able to give that feedback to their product development people because that kind of data started coming up.”
Another topic around use cases that was of great interest (particularly to retailers with high e-commerce sales) was the importance of analyzing digital marketing campaigns and drivers of website traffic. Understanding how to best allocate digital advertising budgets to various channels, initiatives, and campaigns was a very valuable and logical use case.
There was also a marked interest in the topic of personalized offers and personalized marketing. One participant remarked that in their experience, it required bringing together various data sources and applying different analytics to obtain the necessary insights; they also said it required the use of different tech solutions to execute personalized campaigns, all of which could make it challenging to successfully execute consistently and at scale.
5. Hybrid teams consisting of technical and business members increase the success of Data & AI/ML initiatives
Data & Analytics teams provide capabilities that can impact every part of the business. There was discussion about the best structure for the Data & AI/ML function within the organization, should it be:
- Residing within the IT department where they take project requests from across the org
- Embedded in the business side, where the business teams have technical workers dedicated to their projects
- A hybrid model where the Data & AI/ML team independently straddles both worlds
One participant talked about their experience with a hybrid technical/business approach. “In our company, the Data & AI/ML team sits in between the IT and business teams, which I think is the best place because we can speak the business language and we can communicate with IT in a proper manner.”
Another user reinforced that she thinks that data should not be the sole responsibility of the IT team as that might block these new data-driven capabilities from being made rapidly available to all users. “[It’s crucial that] you decipher the data to make it palatable and accessible to the business decision makers that actually consume it but are not necessarily very technical or data-oriented.”
The participants also noted that sometimes AI/ML teams start in one area of the company before transitioning into others. For example, retailers that started with laying the data foundation typically began their journey in the IT organization, but they quickly realized that they needed to include some business perspectives as they began building out dashboards and reports for end users.
Conversely, other retailers noted that sometimes a business team embarks on a project, say, a markdown optimization solution, and eventually realize the necessity of infusing these systems with a continuous stream of granular data—and that usually requires involvement from the IT organization.
Irrespective of where the actual AI/ML team sits within the retailers’ organizational structure, most participants echoed the point that to maximize the chances of success with data &
analytics it is extremely important to include both technical and business perspectives on the team.
The participants also discussed the importance of taking a thoughtful approach to rolling out capabilities to end users to drive usage & adoption. Change management was a key theme of the discussion. Key to this was the need to put the capabilities/insights into the appropriate context for the business end user. It was also noted that it was highly desirable to work towards embedding the capabilities into the end users’ daily business processes, so it is important to incorporate potential business process changes into the rollout planning.
6. To build or buy?
One of the ongoing topics for technical leaders in any industry is whether to build or buy. That is, build a technology that precisely meets the requirements of your organization, or buy one “off-the-shelf” that may have a lot of functionality ready-to-go, but might not offer the flexibility you need.
This is particularly relevant to decisions around Data & AI/ML projects. Particularly when it comes to true cutting-edge applications, off-the-shelf capabilities may not yet available; at the same time, if a retailer wants to build the capability themselves, they may not be able to find (or afford to hire) individuals with the required skillsets.
On top of these considerations, retailers need to take into account a full picture of what is truly required to build out comprehensive Data & AI/ML capabilities and they also need to take into account the operation, maintenance, and updating of any capabilities they build internally, adding a new layer to the challenge that is being addressed.
Some of the considerations and potential approaches that were discussed are highlighted in the following chart:
For one participant, time-to-market was the big decision driver. The described that in instance that they did decide to go with a third-party vendor, they preferred to keep their options open by using short-term licenses so they could transition to a homebrewed solution down the line.
“If there’s enough time, we develop ourselves,” he explained. “But for example, when it came to building a price recommendation tool, we saw that it’ll take more than six months to build, but buying a tool and deploying it would only take two.” He went on to describe how his team went with the off-the-shelf tool but used a one-year renewable license, rather than a long-term engagement and would revisit the question later.