Most services work well for\u00a0the majority\u00a0of\u00a0people \u2013 the\u00a0users who meet the criteria, have the right\u00a0documents\u00a0and understand the process.\u00a0For others, they hit friction points. This can\u00a0show as\u00a0eligibility questions with no room for nuance, payment steps that fail without explanation\u00a0and digital journeys that stop without telling the user why or where to go next\u00a0–\u00a0\u00a0unhappy paths\u00a0through a service that was never designed with them in mind.\u00a0<\/p>\n
Unhappy paths are where\u00a0many\u00a0services fail. They generate failure demand,\u00a0complaints\u00a0and appeals,\u00a0excluding\u00a0the people who most need access. And they are\u00a0almost always\u00a0under-researched because the users who experience them are harder to recruit, harder to schedule and take longer to understand.\u00a0<\/p>\n
Teams, especially those working in Government,\u00a0understand users through ongoing research. In practice, time and cost constraints mean most teams research with small numbers of users on a regular basis, which is rational. It is also how marginalised users end up less visible in the evidence base, and how the difficult edges of a service avoid getting designed properly.<\/p>\n
Working with a national licensing body on their digital service, we\u00a0encountered\u00a0this problem directly. Hundreds of thousands of applications are processed annually, with a valid licence a legal requirement for every role.\u00a0<\/p>\n
The service had\u00a0largely been\u00a0designed around the eligible applicant. What was weaker was the understanding of unhappy paths: applications that failed, applicants found to be ineligible\u00a0and\u00a0edge cases involving complex personal circumstances. These were the users we needed to understand. And they were exactly the users who are hardest to recruit.<\/p>\n
This organisation was able to assemble a larger than usual sample. We researched with over 40 participants, many of whom had declared disabilities, lower literacy levels and complex circumstances including criminal history and right to work ineligibility. The research provided rich, varied data which was exactly what was needed to understand where the service might break down.\u00a0<\/p>\n
We had a two-week window to analyse\u00a0this data\u00a0and generate\u00a0a suite of\u00a0GDS-compliant artefacts.\u00a0<\/p>\n
Faced with 40 interview transcripts and a two-week deadline, the options were\u00a0to\u00a0narrow the sample, simplify the analysis or find a way to do it properly at pace.\u00a0<\/p>\n
Narrowing the sample would have meant losing the very evidence the research was built to capture. Simplifying the analysis would have meant flattening nuance and potentially dismissing difficult accounts as outliers. Neither was\u00a0going to achieve the outcome we needed.\u00a0<\/p>\n
We brought in AI tools to handle the analytical heavy lifting, which is a well-evidenced use of the technology. A 2024 pilot study on the value of generative AI for qualitative research (Pattyn, 2024)(1) found that generative AI completed qualitative coding tasks with four times less effort and fifteen times faster throughput than human coders, with higher inter-coder reliability. Our\u00a0question was\u00a0not whether AI can assist qualitative analysis. It is whether it\u00a0could\u00a0do so without losing rigour or integrity.\u00a0<\/p>\n
Used with discipline, we found the answer\u00a0was\u00a0yes.<\/p>\n
Multi-layered prompts were designed around\u00a0assessment expectations from day one. They functioned as both an analytical method and a form of design documentation. Themes were surfaced systematically across all 40 transcripts, stress tested across research rounds and outputs were assessment-ready by design.\u00a0<\/p>\n
Researcher accountability was non-negotiable throughout. AI outputs were validated against interview notes and human-led synthesis, then workshopped with the\u00a0organisation. Researchers who had been present in sessions verified that generated themes reflected what they had\u00a0actually heard. Acceleration in analysis was only made safe by that human check.\u00a0<\/p>\n
Eligibility was a complex problem in the service.\u00a0A significant proportion of\u00a0\u00a0applicants each year invest in training and pay fees\u00a0despite their\u00a0circumstances meaning\u00a0they will face refusal. Criminal history, overseas checks\u00a0and\u00a0right to work status\u00a0are not simple yes or no questions for many applicants.\u00a0<\/p>\n
An early assumption was that a separate eligibility checker, aligning with established government patterns, would resolve this\u00a0but our\u00a0research\u00a0started to show that\u00a0it\u00a0wouldn\u2019t. Many applicants could not reliably assess their own eligibility, particularly where the determining factors were complex or\u00a0hard to accept. A standalone checker, however well designed, would not catch them.\u00a0<\/p>\n
Because analysis was moving fast enough to feed directly into design exploration, the team did not stop at that finding. Structured workshops examined alternatives against policy requirements, business processes, and user needs. Ideas were tested, found Because analysis was moving fast enough to feed directly into design exploration, the team did not stop at that finding. Structured workshops examined alternatives against policy requirements, business processes, and user needs. Ideas were tested, found insufficient, and replaced. The process\u00a0identified\u00a0embedding adaptive eligibility checks within the application itself as a more promising direction, supporting better decisions at the points that mattered most. That conclusion required the time and space to explore\u00a0which\u00a0AI-assisted analysis\u00a0had\u00a0created.\u00a0\u00a0The team had\u00a0space to\u00a0deliver more robust findings.\u00a0<\/p>\n
The first is about inclusion. Government digital teams aim to include marginalised or excluded users in research, but in practice this evidence often comes from\u00a0a very small\u00a0number of hard-to-recruit sessions. Here, a larger sample made it possible to see where patterns genuinely held and where they broke down. Users who challenged assumptions\u00a0were able to shape\u00a0the direction of the service.\u00a0<\/p>\n
The second is about researcher wellbeing. Much of this research surfaced trauma, distress, and open hostility. Repeated re-immersion in that material is\u00a0a real cost\u00a0to researchers. AI-assisted analysis reduced that exposure while still ensuring those realities informed personas, journeys, and design decisions\u00a0making inclusive research\u00a0more sustainable.\u00a0<\/p>\n
At assessment, the\u00a0panel reflected not only on the volume of evidence but on the quality of thinking it\u00a0demonstrated, particularly where complexity had been worked through in the open. Insights remained traceable to evidence\u00a0and the difficult cases were visible\u00a0and worked through.\u00a0<\/p>\n
Used well,\u00a0we found\u00a0AI did\u00a0not compress research into something less. It\u00a0made\u00a0it possible to go broader without losing depth. In this case, it gave the team the capacity to understand users who had previously been less well understood, and to design a service more likely to work for all of them.\u00a0<\/p>\n
The AI\u00a0tools\u00a0didn\u2019t\u00a0do the\u00a0thinking\u00a0but they did make more thinking possible.,\u00a0<\/p>\n
Want to design services that work better for everyone? Explore our User Centred Design approach, or talk to us about how we can help make your digital services more inclusive, accessible and effective.<\/strong><\/p>\n User Centred Design<\/a> Talk to us<\/a><\/p>\n (1) Pattyn, F. (2024). The value of generative AI for qualitative research: A pilot study. Journal of Data Science and Intelligent Systems, 3(3), 184\u2013191<\/p>\n","protected":false},"excerpt":{"rendered":" Most services work well for\u00a0the majority\u00a0of\u00a0people \u2013 the\u00a0users who meet the criteria, have the right\u00a0documents\u00a0and understand the process.\u00a0For others, they hit friction points. This can\u00a0show as\u00a0eligibility questions with no room for nuance, payment steps that fail without explanation\u00a0and digital journeys that stop without telling the user why or where to go next\u00a0–\u00a0\u00a0unhappy paths\u00a0through a service […]<\/p>\n","protected":false},"author":8,"featured_media":7984,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"content-type":"","inline_featured_image":false,"footnotes":""},"categories":[101],"tags":[],"industry":[],"class_list":["post-8239","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-blog"],"acf":[],"translations":[{"blog_id":1,"post_id":8239,"name":"Europe","code":"EN-GB","hreflang":"en-gb","url":"https:\/\/version1.thinktankdev.org.uk\/blog\/inclusive-research-at-pace-how-ai-helped-us-listen-to-the-users-who-are-hardest-to-reach\/","is_current":false},{"blog_id":13,"post_id":8239,"name":"Americas","code":"EN-US","hreflang":"en-us","url":"https:\/\/version1.thinktankdev.org.uk\/en-us\/blog\/inclusive-research-at-pace-how-ai-helped-us-listen-to-the-users-who-are-hardest-to-reach\/","is_current":true}],"yoast_head":"\n