Why Do We Need AI-by-Design?
Why Do We Need AI-by-Design?
Part 2 of the AI-by-Design series. Start from Part 1.
Why We Talk About AI-by-Design
Ioannis — Head of Global Data Science & AI: "It is easy for data scientists to get stuck in technical challenges, forgetting that we solve human problems. Our solutions will impact human lives in many different areas, from finding a new job to recommending videos to their children. It is our responsibility as data scientists to develop ethical AI."
We view AI-by-Design as an approach to AI innovation that is human-centred, iterative and collaborative. It leverages a designer's approach to problem-solving — building empathy with the user — and applying divergent and convergent thinking to explore and define the problem and solution space.
Three Key Advantages
1. It Takes a Human-Centred Approach
AI-by-Design strives to build a deep understanding of customer needs before developing solutions. This includes recognizing when AI is not the right solution. The team should build whatever solves the problem and empowers humans, instead of prescribing a specific technology.
2. It Leverages Interdisciplinarity
Instead of a cold handover of insights from a design research team to engineers, AI-by-Design encourages cross-silo collaboration. When designers and data scientists work together, they cover each other's blind spots and there is less room for miscommunication.
The "use of design thinking when developing AI tools" has been recognized as the most important differentiator for AI high performers in the State of AI by McKinsey in 2021.
3. Solutions Are Designed for Our Dynamic World
AI models are often trained in a sandbox environment, but they will be used in our messy, complex world. Therefore, AI models need continuous retraining. It is crucial to design feedback loops in which user data and actual behaviour are collected to improve the model.
The 5 Barriers
Over the years, we've noticed that organisations often encounter similar barriers to implementing AI-by-Design:
1. Not Knowing Where to Start
Organisations understand the need to transform but keep approaching AI with the same static mindset of traditional technology — where decision steps are easy to track and humans are in total control.
2. Misalignment Between Technology and the Organisation
When companies identify a new technology, they often see the technology itself as the biggest impediment. So they hire data scientists and developers but don't rethink the company's way of working. The real blocker is organisational culture — how teams are organised, communicate and collaborate.
3. Tech and Design Work in Silos
Teams keep approaching AI with rigid, separated workflows instead of the dynamic, cross-functional collaboration it requires.
4. Underestimating the Importance of Data
Companies often assume that having data means they're capable of AI. In reality, having data doesn't mean you can use it to train an algorithm. Creating high-quality labels is a fundamental step that is often neglected. Some companies even try to develop AI without the groundwork of organising themselves for data gathering. It's like trying to run a car without fuel.
5. Neglecting Trust
It is common for organisations to overlook building trust between the end-user and the AI system. To achieve trust, organisations must be transparent and able to explain how the solution works. Especially since AI promises automation with humans rarely in the loop, it is crucial to develop AI that can be understood.
We believe that this will lead to a more effective way of working, more customer-centric solutions, and will ultimately lead to saving costs and an increase in revenue.
Next: Part 3 — The 6-Step Framework
Previous: Part 1 — Introduction