Building Custom Now Assist Skills in ServiceNow? Token Limits Are Your Hidden Bottleneck.
Perfect prompt engineering means nothing if your responses are getting truncated.
While implementing custom Now Assist Skill Kit solutions, I kept hitting the same wall default token configurations silently breaking real-world GenAI workflows.
Key Backend Tables to Know:
Purpose | Table |
Now Assist skills | sys_one_extend_capability |
Request Token Config | sys_generative_ai_model_config |
Response Token config | sys_generative_ai_config |
This impacts critical use cases like:
. Incident Summarization
· Problem RCA
· HRSD Case Analysis
· AI Agent Workflows
· Multi-record Context Prompts
The real lesson?
GenAI success in ServiceNow isn’t just about what you prompt. it’s about how the model is configured to respond.
Master these 5 pillars:
- Token Optimization
- Context Management
- Response Chunking
- Model Configuration
- Governance & Observability
As Now Assist adoption accelerates, token management is becoming a core skill for every ServiceNow AI engineer not an afterthought.
Have you hit token limits in your GenAI implementations? Drop your experience below
