A simple re-framing of the Kano Model to prioritise AI-powered features.
The Kano Model is a popular tool for prioritising features, though I feel that it needs a slight adaptation when being used for AI-powered features, or really any feature that involves cutting edge technology.
There are many articles on Medium and elsewhere that explain what the Kano Model is, how it works and why you should use it, so if you are unfamiliar I suggest you check those out first.
The main issue I see with using the traditional Kano Model for AI features is that the technology is still so new, many users do not really expect to find it in their products. If you take a product like enterprise resource planning (ERP) software and ask people “How would you feel if the software could identify potential supply chain disruptions and automatically generate emails to suppliers to mitigate these?”, it’s likely that people would not see this as a ‘must-have’ feature, based on what they are used to today. And, since ‘performance’ features essentially mean that a minimum level of that feature is expected, it is unlikely that the AI feature would fall into that category either. I suspect that the majority of AI features would currently fall into the ‘attractive/delighter’ category, which could give the product team the (potentially false) impression that these features might not be worth investing in.
A simple adaptation that should help to address this is to ask people to picture themselves using the product a number of years into the future. “Imagine it’s 2027 and AI has continued to develop at its current rate. How would you feel if your product could [insert AI feature]? How would you feel if it could not?” This simple twist should help people to situate and compare your AI features against their perception of the standard product (i.e., competitors) in your category in the future. It should also provide a sense of people’s expectations on where the technology will be in a few years and help you to plan accordingly.
Don’t forget that doing this also requires the response options to be slightly adapted. Rather than the usual “I like it”, “I expect it” and so on, these should become future-tense; “I would like it”, “I would expect it”, etc. The actual analysis and graph plotting remains the same as traditional Kano.
Finally, in order to do this type of study, it’s important to recruit participants that not only fit your current profile, but that also have some level of experience or familiarity with AI already (perhaps from playing around with ChatGPT or other AI-powered products). This is also an opportunity to include some open-ended questions around use cases, concerns, AI trust and anything else that might be useful to your team’s development efforts after you’ve advised them on which features to prioritise.
Happy researching! And if you try this method (or have tried other adaptations of Kano), be sure to let me know :)
