AI Search Best Practice for Production Implementation

[Paul Curwen]

Just started my journey with AI Search. Anyone care to share their experience implementing AI Search in production.

Some initial basic questions: 

1. Did you implement directly in production using Guided Setup, or use update sets from sub-production etc ?

2. If you have multiple Portals is it possible to put AI Search on just one of those and leave the other as Zing?

3. How much effort did you put into post-go-live tuning, and what amount of regular tuning effort have you found is required?

Any gotchas much appreciated 🙂

[Gerard Dwan]

Hi Paul, great question! I'm hoping to hear more from customers on your question, but as a starting point: 

1) Most customers use update sets after testing and validating in a pre-prod environment 

2) You enable AI Search on your instance, which makes it available to use, then you enable it only on the portals you would like to use it. It will not replace or override anything that you don't explicitly configure. 

3) Assuming there are not significant content changes, we see customers have an initial spike of activity as they analyze user behavior and determine the initial setup for items like Result Improvement Rules or refinements on synonyms.

The benefit of AI Search is that the relevancy is tuned automatically and monthly, so we see ongoing efforts about a half of a day per month to review analytics and make tuning adjustments. We have this article that articulates some of the go-live and tuning activities for context: Going Live with AI Search? Follow these steps <- that could help with some additional best practices. 

A couple of gotchas,

• this article talks about the steps you may need to take if you are enabling AI Search in a portal that has had the search widgets customized. 
• long running user criteria will make search slower, so it's something to look at in a lower environment. Test searches with non-admin users!