.png)
In the last blog, I explored user-centred design for a product where the solution had been defined but the user's problems hadn't. Reflecting on the process of building trust, we got to know our users and their needs and were transparent with the process. Consequently, we were able to unearth potential barriers to successfully adopting the new AI tool. A lot of focus is on the technology, but the failure can be due to not recognising how users want to use the technology: implementation.
In the case study below, we delve into this more. The AI solution was built in parallel with product development. This led to product discovery informing about AI development and exploring ethical issues with users.
Problem Statement
Primary care practices are overwhelmed with patient requests, leading to challenges in triaging and helping patients get the care they need.
Research Question: How could AI be used to support administrative staff in triaging patient requests and improve the overall patient experience?
The intention was to support primary care staff with triaging the seeming deluge of patient requests at their practice. Here, the assumption is that AI will support admin services in triaging. Why were patients struggling to get the care they needed and how could we support staff to manage patient requests? Do we necessarily need AI?
Problem exploration
User Research: Interviews with GP staff, patients in waiting rooms, and sitting in Patient Participant Groups.
Advisory Group: Set up an Advisory Group with patients, GPs and admins to provide in-depth insights and address ethical considerations.
User Journey Mapping and Validation: Analysis and mapping of user journeys to understand the current process and potential pain points.
Quantitative Analysis: Evaluation of user engagement to identify areas for improvement.
Key Findings
Patient Frustration: Patients felt disempowered and viewed appointments as a primary means of obtaining answers. Existing digital front-door services (where patients access GP services through an app) were showing some success. However, in some ways this was proving to be a victim of its success as some practices set up long lists of services for patients to pick through, leading to the digital version of ‘Press 1 for help’ feeling. This was intimidating for those not familiar with the health service or where English is a second language.
Admin staff (inc receptionists): Manual and semi-automated pathways sorting requests to the correct personnel exist, but it can be tricky for staff to manage complex requests with limited context.
Clinicians (GPs, prescribers, nurses): AI could be a big help in deciding what to do next for different patients. Clinicians often have to wade through patient notes to make sure they can make the right decision for their patients and treat the patients how they need to be treated.
Conclusion
After digging deep into what patients and doctors were going through, we realized that the current system wasn't exactly cutting it. People were getting frustrated and having a hard time navigating all the options.
That's where our idea came in: We created a simpler way to search for information. We were also finding a way to test the user experience whilst developing the technical stuff mindfully in the background. Win-win in designing the product in parallel with the AI model I'd say: building both parts in parallel meant we could feed into the design of the product and AI model to create a final solution that would improve overall users' experience.
Co-designing with end-users meant they were able to highlight issues that the team hadn’t considered. This resulted in a much more robust pathway that didn’t lose the nuance of patient care. It also was clear to us that AI didn't need to solve everything despite the hype!
A more human-centred design mindset could help to recognise the complexities in this wider infrastructure. Introducing AI products in healthcare isn't simple. The user-centric design approach highlighted the direct users' needs which we can design for but for the AI product to be a success, you need to understand the context in which the product will sit: indirect users buying the final product for their practice, hospital, county or even country. That's a product thing to do: take over the world!