Generative AI and Padel

Real-world testing has always been vital with any new technology.

The excitement surrounding Artificial Intelligence’s (AI's) potential often leads to a ‘think big, act big’ mentality. While ambitious thinking is essential, it must not overshadow the importance of actionable insights, user-centric design, and real-world applicability.

The pace of technological change is relentless, and the stakes are high. Organisations must balance the need to establish foundational data science and digital capabilities with the agility to identify and seize opportunities for learning from real customers. The mantra should be clear: think big, start small, test early, and test often. By focusing on desirability before feasibility, we can create innovations that solve real-world business problems and meet user needs.

Why real-world testing is even more important with AI

Real-world experimentation with AI is not just a nice-to-have concept; it's a practical necessity. The transition from lab environments to real-world applications has exposed discrepancies in performance, limitations in training data, and complexities in real-world settings. These challenges have drawn attention to biases in AI and emphasised the importance of continuous testing and deployment.

It's within this complex landscape that our collaboration with the Lawn Tennis Association (LTA) emerged. The project, aimed at enhancing the game of padel through generative AI, serves as a tangible example of how real-world experimentation can lead to innovation, learning, and value creation.

Going from idea to reality in 6 weeks

Padel, an exciting fusion of tennis and squash, has been growing in popularity. Over a 6-week design sprint with the LTA, we were challenged with making this sport more appealing and accessible to Gen Z. Gen Z are commonly defined as those born between the late 1990s and the early 2010s. We embarked on first-hand research, co-creation, and deep dive conversations with coaches, players, and over 20 Gen Z-ers themselves.

Through this, we unearthed four pivotal insights that would guide our project. Firstly, Gen Z wanted narrative control — a way to actively co-create their experiences rather than passively consume them. Secondly, we discovered their desire for quick and user-friendly experiences — technology that was intuitive, seamless, and instantly gratifying. Authentic personalisation emerged as the third insight — a call for non-generic, tailored experiences that respected individuality. Lastly, given their digital prowess, Gen Z displayed an intrinsic curiosity about generative AI, and a desire to comprehend the technology, not just use it.

We then co-created specific AI-driven experiences, including:

• AI photo booth: Designed to satisfy the need for narrative control and personalisation, this photo booth allowed players to edit, share, and augment their photos using open source AI image generation models.
• Virtual courtside instructor: This AI instructor served as an interactive guide, providing game tips, educating players about the sport's history, and answering specific questions using an AI text generation model.

Figure 1: The AI photo booth (left) and virtual courtside instructor (right)

Real-world testing and learning

We deployed these prototypes in a live experiment as part of a 3-week pop-up event at the Stratford Padel Club, with the aim of gathering user feedback about the ideas and understanding the limitations and opportunities for the future. This process often leads to features being refined, dropped, or re-imagined.

As anticipated, the AI's interpretations were not flawless. We encountered issues with false answers from AI and imperfect images in the photo booth. We were particularly confronted with biases in the photo booth, especially concerning gender and race representation. Though we were aware that these issues might arise, witnessing these shortcomings first-hand reinforced a vital truth: real-world implementations are intricate and unpredictable, and this is where some of the key challenges and limitations of generative AI become apparent.

Conclusion: Building a positive future with real-world insights

Our experience with the LTA project underscores the importance of real-world experimentation with AI. By thinking big but starting small, testing early and often, and putting users at the heart of the process, we were able to create value and learn invaluable lessons.

New technologies must evolve within the context of human society, which is diverse, complex, and continually changing. In dealing with people and the extensive datasets that train generative AI models, the importance of early and frequent testing has never been more pronounced. To align AI with its intended purpose, we must engage with the very people it's designed to serve as early as possible, absorbing their feedback to shape an experience that resonates with their collective needs and aspirations.

It’s vital that all organisations, both public and private, understand, test, and deploy generative AI with real users and do so safely and responsibly. We all have a role to create the positive and responsible AI future that we all collectively want – one that enables human creativity, boosts engagement, and fosters inclusivity.

Our work with the LTA shows how Deloitte’s decades of experience in bringing together human insights with novel technologies can help make a brand and product more accessible, engaging, and exciting.

If you are interested in finding out more about any of our experiences and services in this area, please do get in touch.