Just Because You Can Doesn't Mean You Should

AI and predictive design is very popular right now, but relies on a very large, historical data set to be successful. ​I work with a diverse set of clients across an array of industries and find myself having the same conversation regarding including AI in new products.  

Client:  We are super excited about AI and want to leverage it in our product.
Me:  Great, tell me about your idea.
Client:  We are going to use AI to remove all friction and create a one touch experience that will be amazing.
Me:  Ok, how have you identified your friction points?
Client:  Well, we do not have solid data, but we are not seeing the conversions we expect so we know they are there. We have to make the experience seamless.
Me:  Ok, what does seamless mean to your consumers?
Client: We use AI to automagically do everything on their behalf so they just have to click buy.  
Me:  Are you sure?  That sounds like what seamless means to you, what does it mean to your consumers?
Client: ...
Me:  Do you have enough data to reliably execute the experience?  Will the experience work for at least 80% of your consumers?  Are you introducing unnecessary opportunities for error? Is this something your consumers actually want?
Client: ...
Me:  Let's take another look at this from your consumers perspective and see how we can better facilitate their experience in the near term while gathering data.

Most new products simply do not have enough historical data to create accurate training data for a specific problem. Anticipatory or predictive design is intended to make decisions on the behalf of the consumer to reduce cognitive load.  The wrong decision creates friction and overall dissatisfaction with the product.  Until you have enough data, it may be better to have your consumers tell you what they want rather than try to guess their intent.

Just because you are able to execute the shiny new thing it doesn't mean you should.  Take a step back and take time to gather data.  Once you have enough data to reliably understand your consumer's  behavior you can incorporate predictive design  practices successfully.