How to implement ServiceNow's Generative AI to get solutions for incidents?

[Navaneeth1]

I have a ServiceNow Custom Scoped application that implements Incident and Change Management functionalities by inheriting those tables as a new scoped application table. 

I have a requirement where the clients would like to see the solution of incident tickets based (and sometimes even suggestions) on Implementing AI to learn from older tickets, kb articles, and Question and Answer feedback from clients which were recorded on the additional comments of the incident form.

I read about ServiceNow's Generative AI and NOW Assist but I don't know how those can be used to analyze old ticket information to get and suggest precise solutions for new tickets. Can someone please help me by providing some information on how these AI Capabilities can be implemented for my requirements? 

[Stefan3000]

Hi,

The only solution I know of is to create a Vector DB (out of the KB and Tickets), for example in Azure ML Services, and build a Retrieval Augmented Generation Prompt Flow, then use the API in ServiceNow.

However, I'm highly interested in hearing about alternatives and utilizing more features already included in ServiceNow.

Greetings,

Stefan

[Its_Azar]

Hi there @Navaneeth1 

Thats a great question and guess we can use NOW Assist, to provide solutions and suggestions for incident tickets based on historical data. Here's a general approach to implementing this:

First, ensure that NOW Assist is enabled in your ServiceNow instance. 

Gather historical data from incident tickets, knowledge base articles, and client feedback stored in the additional comments field of the incident form. make sure that this data is structured and labeled appropriately for training the AI model.

Preprocess the collected data to remove noise, standardize formats, and extract relevant features. 

Train the Generative AI model using the prepared data. 

Integrate the trained Generative AI model with your incident management process. by configuring triggers or business rules to automatically generate solutions or suggestions based on the content of new incident tickets.

If this helps kindly accept the answer thanks much.

[Navaneeth1]

@Its_Azar 

Thank you so much for this response! Do you know how I can train the Generative AI Model with the incident data?

[Miguel Donayre]

Hi,

There is a huge misunderstanding about how the Now LLM, which powers our Now Assist products. You can not train it on your own use case, but it doesn't work as if I don't take your transaction data to train it. The Now LLM isn't trained like our other AI products. 
You can use the Gen AI Controller to create a customer integration with Open AI, Aruze, or Google's AI model to train for just your use case. Just keep in mind it will be completely custom. It was a suggestion in the answers.
Other AI products were mentioned in the response, like Agent Assist and Predictive Intel, that can come close to helping your use case. 
We do have a use case and "Skill" that Now Assist offers, and we are continually expanding those capabilities,

However, what you suggest as a solution is not possible and will not work. 

Here is the link to all Now Assist capabilities so you can understand the capabilities of the Now Assist - https://docs.servicenow.com/bundle/washingtondc-intelligent-experiences/page/administer/now-assist-p...

I hope that clears things up!