Backend vs Frontend Taxonomy

[drosenst2]

In your experience, how do you all differentiate the browsing/filtering structure for the backend vs. the frontend taxonomy for knowledge articles? Is it exactly the same or different? Do you match that structure in other HR/IT applications across the board outside of SNOW? 

Also, do you see a big difference categorizing every article to help with Now Assist/Chatbots/AI in general?

I have a lot of thoughts and experience doing this, but don't want to bias the conversation. Any feedback is greatly appreciated.

[David Kay]

Ideally, all the taxonomies in your enterprise will map to one another.  That is, they need not be identical, but each one should be a subset of the others.  For example, the product taxonomy on a case or an incident might need to be three layers deep.  If the taxo used in knowledge is only two layers deep, it should map to the top layers of the incident taxonomy.  In fact, I'd do my best to avoid separate front-end and back-end taxonomies.

I recently lead a taxonomy workshop that included the CFO, because he wisely wanted to make sure that his reports aligned with how the rest of the business talked about product lines and products.

Here are some more detailed thoughts, one of a three-part series: https://www.dbkay.com/knowledge-representation/what-makes-a-good-taxonomy

[Kohei Tominaga1]

Hi, @drosenst2 

I will share my idea.

I think these Backend and Frontend taxonomy should be treated differently.

• End users tend to classify information based on their general knowledge and past experience
• Fulfillers (agents/content owners) tend to classify based on organizational structure, services, or business processes

Because of this difference, having separate taxonomies would be ideal.

However, in ServiceNow knowledge management, we usually cannot dynamically present different taxonomies per user type.

So historically, I believed:

• We should prioritize user-facing taxonomy
• Because the number of end users is much larger

This assumption is now changing.

Modern AI (Now Assist):

• Does not rely on category structures
• Instead, it uses content relevance and semantic understanding

Because of this:

• The importance of user-friendly browsing taxonomy is decreasing
• Users increasingly rely on search / conversational interfaces rather than navigation

Given this shift, I would suggest:

• It is reasonable to prioritize a taxonomy that is easy for fulfillers to manage and maintain
• Instead of forcing a user-centric structure that may not reflect how content is actually maintained

If AI capabilities continue to expand:

• Even fulfiller-oriented categorization may become less important
• Knowledge may be retrieved primarily via semantic search without relying on explicit classification

[drosenst2]

I was going to ask about Now Assist and my way of setting up knowledge is different. We are putting a huge focus on meta, but is Now Assist OOTB NOT looking at backend categories and frontend user-facing taxonomy? Are they only really used for search browsing and filtering Now Assist results? 

[David Kay]

Without a user-facing taxonomy, you're likely shortchanging other AI models assuming you're ever going to use a different system (which is a safe bet, as early on as we are.)  Also, do you really want to preclude users from ever searching or doing RAG for (say) a particular product?