Let’s get one thing straight: AI will never replace user research. It just won’t. Now before you start applauding my brave stance, know that we are very pro-AI. Divinate is a tool that uses AI for helping you make sense of user research. We like AI.
I’m all for leveraging cutting-edge technology to make our lives easier and our work more efficient. Lest we forget, one of the main things technology does is replace work. And that’s okay—we always find things to do. Some things, though, shouldn't be automated because the act itself has value.
Talking to your customers in one of those acts. It inherently has value and should not be replaced. I'm going to defend that position here. Buckle up.
Can AI just do research for me?
No. Alright see ya later! Okay but really, let's dig into why this isn't a thing, and why you shouldn't even try to replace talking to your customers with AI.
Replacing the customer?
AI users will never pay for your product. I don't really care what your synthetic users had to say. Call me when GPT is pulling out the credit card. You can't replace the customer in this equation. That's potentially the most important variable.
AI is trained on existing datasets. It is inherently, by it's very nature, not good at creating new data. So when you're asking a chatbot what it thinks about your product it is generalized and taking guesses, not actually basing responses on reality. That's the definition of biased.
Biased data only serves to make you feel good about yourself. That's not how you win in the market. Remember, research isn't a box to check off. Research is supposed to help us make product decisions.
Replacing the researcher?
You can run an AI automated research process all day, but the real value in research is face to face, human to human conversations. Having a non-human represent your company is not only odd and impersonal, but a bad brand look. You can't replace the researcher.
Ultimately, the data you're collecting in this application is just a worse, non-deterministic survey. There is plenty of room to use AI to make some cool survey experiences, but you need to remember that it's exactly that; a survey.
If you reach saturation at 11 conversations, how much effort did you really save automating thousands of conversations? To me, it feels more like you've sold out the best part of conducting research in favor of checking off a box in less time. That's the definition of the middle-range research problem.
What are we doing here?
Why would you want to remove yourself or the actual human on the other end from that conversation?
First ask yourself what user research really involves. Talking to humans. The core of research is asking the right questions, listening to the answers, and then digging deeper to understand the underlying motivations.
The value of user research is in the human stories we uncover. Those stories are rich and powerful. They can't be automated away. I've seen hard-headed stakeholders who were nearing customer-hostile change into customer advocates because of the stories we've learned in research. It's transformative.
We, ideally, should be learning something unexpected when we conduct research. User research is also centered around empathy building. You want to know what builds a ton of empathy? Sitting and talking to your customers for 30 minutes to an hour.
AI can process data at speeds and scales that humans can’t match, but it doesn't really understand the context outside of what it's told. Obviously AI can’t experience emotions (yet?), it doesn't know why past decisions were made, and it certainly can’t build the kind of empathy that’s crucial for genuine user research. (And again, it's not going to pay for my product.)
Augment, not replace
Now, to be clear, there are many parts of the process that are time consuming and laborious for no good reason. Writing research guides can take a long time–AI can help speed us up here.
Synthesis takes way too long. Some of the physicality of moving sticky notes around can be useful to help you think, and if it does, that's awesome. However, 90% of the work of sorting and organizing the raw data can and should be automated.
My bias is to get past the blank page problem as fast as possible and into the parts where my personal touches matter most–conducting the research, applying it to a strategy, and designing something from it.
Using AI to sort, organize, parse, summarize, splice, slice and dice, combine, compare, and mash data together will be a part of the research process from here on out. It's not going anywhere. The teams that will rocket ship ahead will be the ones doing that while also still talking to humans as humans. Ride that wave.
We’ve spent a lot of time thinking about how AI can complement the research process rather than trying to take it over. Our approach is to use AI as a tool that empowers researchers to work smarter, not harder.
Here’s how we see AI adding value to user research
1. Making sense of a lot of data quickly. AI can sift through massive amounts of data quickly, identifying patterns and trends that might take a human researcher weeks to uncover.
2. Finding connections you've missed. AI can cross-reference data from different sources, highlighting connections that might not be immediately obvious. This can help researchers uncover insights that they might have missed otherwise.
3. Streamlining repetitive tasks. AI can automate the more mundane aspects of research, such as transcription, tagging, and basic analysis. This frees up researchers, designers, and product people to do what they do best.
The Future of AI in User Research
We’re excited about the future of user research because we see it as a collaborative effort between humans and AI. We’re not interested in replacing researchers. We want more research conducted more often. More human stories.
At Divinate, we’re building tools that enhance the research process without losing sight of what makes it valuable in the first place. AI won’t replace user research, but it will make us faster at making sense of it. Research is a competitive advantage. Use it.