So, how is the System Concepts UX team exploring AI?
This article brings together some of our thoughts, experiments and examples. It may feel slightly like a mishmash, but that reflects how we are currently exploring what AI means for our work. AI is a classic example of “you don’t know what you don’t know”.
Ethnographic research has always been one of my favourite types of project where you keep broad set of objectives rather than having specific research questions, because if you don’t know what you don’t know, how can you ask specific questions? You need to go with the flow and remain open to what you discover.
The release of Alexa was another classic example of “you don’t know what you don’t know”. Voice-assistant software doesn’t present a menu, lay out all the options or clearly display every feature it offers. If you were not comfortable experimenting and simply ‘seeing what happens’, you were likely to use only a limited range of voice interactions.
AI is similar. How do you know what it can do when there is no clear list of options — especially when that invisible list changes, updates and expands faster than you can keep up? That is why, at System Concepts, our experience of learning how AI can support UX research has not been neat or linear. We are all learning, experimenting, sharing knowledge, getting frustrated, identifying errors and being pleasantly surprised, often all at the same time.
It has been a fun, worrying, impressive, forward-looking, collaborative and technical time. You could even say, “it’s been emotional” — in the style of Vinnie Jones in Lock, Stock, if you are old enough to remember the reference. My sons are 12 and 14, and I find it hard to look at AI without feeling both excited by its possibilities and concerned about its potential impact on the job market for young people over the coming years.
So, where are we now in terms of AI as a UX team? First came the research — and, even if I do say so myself, research is something we are pretty good at. We created Miro boards, read articles and thought pieces, noted technical specifications, and used AI to ask multiple questions about AI.
Because the landscape is changing constantly, this research is ongoing: the Miro board continues to expand, and we keep sharing what we learn. However, two prerequisites quickly became clear before we could begin using AI to support our work:
If we put any data into an AI tool, we need to be confident that it is secure, GDPR-compliant and will not use the information to train its models. That meant identifying the most appropriate tools and ensuring that they offered enterprise-level security.
We also needed an AI policy before anyone began using the technology. It had to set clear guidelines and guardrails covering client confidentiality, contractual requirements, personally identifiable information and the quality of our work. Our principle is that we will never use AI to produce our work; we will use it only to support and add value to work led by skilled researchers.
With these foundations in place, we can now experiment with AI as a sounding board, proofreader or assistant. We were quick to establish that it is no substitute for a skilled and experienced researcher: it exaggerates, displays more confidence than its answers justify, misses nuance and has a clear case of confirmation bias. However, AI can add genuine value for us and our clients, where client contracts allow. Our qualitative work is often guided by specific objectives and research questions. When we focus on answering them, valuable insights beyond the agreed scope can end up on the cutting-room floor. AI can help us gather those clues — indications of where further research could focus or where entirely new insights may be emerging from another “you don’t know what you don’t know” moment.
AI can act as a thought partner. For example, a researcher might ask: “These are the key themes I have identified — have I missed anything?” It can also automate narrow, routine tasks such as refining wording, producing transcripts and summarising relevant past research. Ultimately, AI adds the most value when it tests our researchers’ thinking, not when it attempts to replace it.
So, that was a brief tour of our AI rollercoaster that we are still currently riding. Luckily, once you are aboard the rollercoaster, it doesn’t feel as daunting as when you are standing on the ground watching it whizz past.
We offer a range or UX and accessibility research, if there is anything our expert researchers can help you with, please feel free to get in touch by clicking on the 'Get in touch' button.
