I was lead by a lot of other references to read up on the “Search to Service” concept. I happened upon this article by Forerunner Ventures making the case that Search is Being Reborn as Service.
They make the now ubiquitous case that “the next evolution of search won’t look much like search at all. Rather, discovery will take the shape of tailored, dynamic services, where users can not only access information but interact with a highly personalized, actionable edit of what’s uniquely relevant to them.”
Yes! AI is transforming user/consumer behavior. But as an ecommerce professional, how does this impact what you do to sell products?
In these articles, The Search to Service concept is entirely cast through the lens of the user/consumer, it says nothing about the mechanics (engineering) required to get AI to sell your product.
Taking a look at employers job openings, they speak about hires that can engineer “AI-powered discovery experiences” into the products or services of the business.
What are AI-powered discovery experiences?
Tell me if Im wrong, but I dont see how this is any different from old fashioned Search Engine Optimization using the now 5 year old Question and Answer optimization that is still required to get into the PPA (People Also Asked) results.
Lets take a look DayDream.ing which is funded by Forerunner Ventures. (I make no judgement about the investment. It might just be what people want. And I wish them well.) The DayDream.ing interface provides the user with example prompts: “Find me a leather bag for fall under $300.”
So from a technical engineering “optimization” perspective, how do we get the AI to provide “the products” that the user is looking for? The AI cannot derive which products fit this PROMPT (search) without content that contains the clues for the AI to utilize. Those clues are search terms and keywords.
The product content must be optimized with the CLUES (keywords); fall, leather, bag, price. The AI cannot function without those clues built-in otherwise. But honestly, what else can you do but anticipate the possible uses of the product and embed them in the content – just like the old Q&A format?
If you ask AI how to bake a cake, it still requires that someone posted “how to bake a cake” online. And that recipe contains ingredients and instructions, which AI will happily provide in a chatty way.
Why not ask AI to “Bake a tunsgten cake”? Since there is not such recipe published online, AI will resort to chatty “creativity”, but there will be no citation for “product” on offer, because the content simply does not exist.
Perhaps I could write more about this, but I think would be beating a dead horse. If you happened to be alert when Google got this patent about “Selecting answer spans from electronic documents using neural networks“, in 2020, you could see the shape of things to come.
From an technical engineering point of view, what you need to do to develop “AI-powered discovery experiences” into a products or services is well established practice that you should have mastered 5 years ago, and 3 years before there was a ChatGPT.