Prototyping public services with AI – the designer’s new role: a human in the loop
AI is an accelerator for designers in prototyping, not a replacement of them
(Part 2)
In my last post, I discussed the advantages of and practical tips for prototyping with AI. Here I’d like to share our thoughts on the role of interaction designers in this process.
While we have been benefiting from the superpower of AI, we are aware that this has changed our role as designers in prototyping. But AI is an accelerator for designers in prototyping, not a replacement of them.
We still need designers to make design decisions, at least for the foreseeable future.
AI is enough for prototyping, but not for production
The main purpose of prototyping is to create a piece of work that allows delivery teams to:
- put something in front of users for usability testing
- communicate complicated logic to developers
AI is powerful enough to deliver imperfect prototypes to fulfil the above purposes, when we intend to throw them away afterwards. However, the quality of the code, including its clarity and structure, is not yet suitable for production. Developers’ involvement is still needed if we want to build robust and accessible services.
Collaboration between developers and designers remains a key activity in our agile development process.
Designers can make design decisions that AI cannot
A designer’s job is to provide design decisions based on evidence and user research findings. AI is not ready for that – it’s unable to judge what factors will contribute to improving accessibility on a usability and semantic level. Designers and researchers have contextual knowledge about the services they are building. When they see users being hesitant in scrolling down a page, they will likely know what contributed to that hesitation. AI doesn’t.
For the interaction designer, AI handles how to code a solution, but the designer remains essential for determining how the solution is based on user insight.
Context and domain knowledge: research is needed
AI struggles when tackling services based on new policy or internal domain knowledge that isn’t publicly available. For example, AI could have a good attempt at responding to this prompt: Create a prototype that asks drivers a series of questions as part of the process of applying for a driving license
Using readily available information, the output would likely be impressive.
However, it would struggle significantly with a prompt like this: Create a prototype that allows delivery officers at DfE to manage new targeted support for schools, including a way of monitoring progress and a process to handle when schools do not engage as expected
The latter example is based on an entirely new policy for internal users, requiring user research to establish and design the correct service flow.
AI can help make a start, but human designers with contextual knowledge are still very much needed to determine the service logic and user needs.
AI could be a good collaborator but not a lead designer
Sometimes with design, usability testing between several versions can be a great way of establishing user needs and preferences. Cursor AI could absolutely help speed up the process of mocking up variations on a theme, and perhaps even suggesting pros and cons of different versions. But for reasons apparent in the reflection above, AI might not be so good (yet!) at establishing what the theme might need to be. It might also not be so good at eliciting from users what their preferences are, and why. This is where the designer’s qualitative research skills – observing body language, listening for hesitation, and probing for ‘why’ – remain irreplaceable.
AI empowers the designer to execute variations faster, but the designer is the one who sets the research agenda and interprets the findings.
Elevating the designer’s role
By automating the repetitive and complex coding tasks, AI effectively elevates the interaction designer from a ‘prototyper’ to a ‘strategic problem solver‘.
Our focus shifts to the higher-value activities: understanding user intent, mapping complex service journeys, and designing effective content flows. This means more time spent on true design thinking, leading to better public services overall.