Thales

The Opportunity

Thales is a multinational aerospace company that specialises in Artificial Intelligence (AI). We had the opportunity to help them gain insight into the human decision-making process. There was a specific focus on how AI can enhance decision-making in the context of grocery shopping.

Our Approach

I consciously elected to take a human-centric approach throughout the project, because I believe to make better products and services, one must design with the end-user in mind. Therefore, I pushed the team to interview as many people as possible, with the hope of gaining insights into how people shop, and how they make decisions.

After interviewing people about their shopping experience, we created journey maps that would convey the key parts of each person's decision-making process. Using these journey maps, we applied and evaluated three human-decision making models (OODA loop, Vroom-Yetton-Jago and Recognition Primed Model). From this, we found that the Recognition Primed Model was most relevant, although we did incorporate parts of the other models in our final solution.

Following more interviews we felt we had enough information to start creating personas. Each persona acted as a distinct user group that we could then specifically target with any final solution. From a design-process perspective, this accelerated our ideation phase, as we had concrete user-groups to think about when presented with any idea.

I undertook a review on the current literature surrounding Artificial Intelligence and I carefully evaluated the realistic options we had as a team to implement. This was key to our success, as any solution depended on the feasibility and effectiveness of techniques in the field of Artificial Intelligence.

The Solution

From my secondary research, I knew well-trained AI could outperform humans on decisions that rely upon analysis rather than intuition. Our primary research suggested that most of the shopping decisions are (initially) analysis based. I then suggested we use a technique called Collaborative Filtering to enable the AI behind our solution to make accurate suggestions on limited data. I found this could be implemented with neural network & machine learning techniques.

Our final solution was two-fold. Firstly, we delivered a report to Thales containing our model of human decision-making in the context of grocery shopping. As the technical lead in our group, I identified how and where AI can be applied.
Secondly, we suggested one possible solution where AI could be used: an app that streamlines the grocery shopping experience. Once you have created or scanned a shopping list into the app, the AI will suggest the best shop to go to based on distance, availability of items and cost. If items weren’t available, alternatives would be suggested.

We tested this concept using paper prototypes and got mainly positive feedback. After iterating multiple times, we then presented our final concept to Thales. My role in the presentation was to explain complex technical jargon and algorithms in a clear, simple and easy to understand way.