Introduction
As developers and admins, we’ve all been there: you’re reviewing an Update Set with 50+ customer updates, and you’re trying to piece together exactly what was built. Reviewing XML entries one by one is time-consuming and error-prone.
To solve this, I’ve implemented an Use Case Now Assist "Update Set Summarizer." This tool uses GenAI to analyze the technical "Customer Updates" (sys_update_xml) records and transform them into a human-readable, simplified summary. Here is the full guide on how I built it.
The Workflow Overview :
The solution consists of four primary components:
- Skill Input : Defines the input and deployment.
- Subflow (Tool Editor) : Gathers and cleans the technical data.
- Prompt : Instructs the LLM on how to interpret and format the summary.
- UI Action & UI Page (after Deployement) : Provides a seamless user experience on the sys_update_set form
Step 1 - Defining the Skill and Input
First, I created a custom skill and added a skill input record of the update set record. This ensures the AI knows which specific update set it needs to analyze.
Step 2 - Prompt
Then I added a prompt for generating the summary of update set's customer updates record. I used the Prompt Editor to define a persona and strict formatting rules.
Prompt Content :
Role: Senior ServiceNow Architect.
Objective: Transform the technical subflow payload into a clean, line-by-line summary.
Formatting Instructions:
For every record in the input, provide a descriptive summary using this exact structure:
[Artifact Label/Name] ([Type])
Action: [Simplified Action]
Summary: [Functional Description]
(Note: Ensure the three asterisks *** are used between every single record to create a
visual separator line.)
Data Input: {{GetCustomerUpdates.payload}}
Update set name : {{update_set.name}}
Strict Processing Rules:
Visual Separation: Place a horizontal line (--- or ***) after every single update
summary.
Translate Jargon: * Replace INSERT_OR_UPDATE with "Created or Modified".
Replace technical types like sys_one_extend with "AI Integration Framework".
Replace sys_gen_ai with "Generative AI Configuration".
No Sys_IDs: If the Label is a sys_id, ignore it and use a friendly name derived from the
Name field.
Detailing: If the record has a 'Description' in the input (like the Business Rule),
incorporate that into the Summary.
Step 3 - Tool Editor & Subflow Logic
Raw XML logs can be messy. To make the data LLM-ready, I created a Subflow in the Tool Editor named "GetCustomerUpdates". This subflow takes the Update Set sys_id, iterates through the associated sys_update_xml records, and returns a clean string payload.
Example Raw Payload:
Label: Demo Table | Name: sys_app_application_f8d... | Type: Application Menu | Action: INSERT_OR_UPDATE
-> Subflow (Input to Prompt):
-> Action Logic
SubFlow Output :
Label: u_demo_table_user | Name: sys_user_role_70d4249xxxxxxxxxxxx1a21f1bdf576 | Type:
Role | Action: INSERT_OR_UPDATE | Description: Role
Label: Demo Table | Name: sys_app_application_f8d4249xxxxxxxxxxxx1a21f1bdf57c | Type:
Application Menu | Action: INSERT_OR_UPDATE | Description: Table creation application
Label: Demo Tables | Name: sys_app_module_01d4249xxxxxxxxxxx21f1bdf57f | Type: Module
| Action: INSERT_OR_UPDATE | Description: Table creation module
Label: Demo Table | Name: sys_ui_application_41d42491xxxxxxxxxxxx21f1bdf581 | Type:
Application Menu | Action: INSERT_OR_UPDATE | Description:
-> Using the SubFlow In Now Assist Skill Kit
Step 4 - Deployment & Skill Settings
Step 5: Frontend Implementation (UI Action & UI Page)
This all I had deployed as UI ACTION on UPDATE SET (sys_update_set) table using the UI PAGE
which basically displays the Structured response given by now assist.
Conclusion :
This Update Set Summarizer skill kit is a perfect example of how ServiceNow Now Assist can be extended to handle technical platform artifacts. By leveraging Now Assist, we've turned a manual review process into a 1.5-second automated summary. This not only improves productivity but ensures that documentation is accurate and easy for stakeholders to understand.
Note : For full Implementation Guide and all the Scripts Refer the Below attached pdf.
