Who This Is For / When to Use
Use this update to:
Reduce AI usage costs across workflows
Scale AI-driven automations more efficiently
Improve performance of high-volume workflows
Run longer or more complex AI processes within the same budget
What Changed
The AI Agent action has been optimized to reduce unnecessary token usage at every stage of execution.
Cleaner Context Handling
Removed duplicate and unnecessary data from prompts
Excludes raw database internals and redundant contact fields
Reduces payload size sent to the model
Smarter Tool Responses
Tool outputs now include only relevant fields
Example: Instead of dozens of fields, only key data like:
Name
Email
Phone
Eliminates excess data being passed into the model
Conversation Memory Optimization
Older conversation steps are automatically summarized
Recent messages remain in full detail
Prevents sending full conversation history on every execution
Structured Output Improvements
Final outputs no longer duplicate full context
Reduces token-heavy extraction steps
How It Works
The system automatically optimizes token usage during AI Agent execution.
Workflow triggers the AI Agent action
System filters and compresses context data
Tool responses are minimized to essential fields
Older conversation data is summarized
AI generates output using optimized input
Performance Improvements
Measured on identical workflows with no loss in output quality:
First LLM call: ~36% fewer tokens
Total execution: ~20% fewer tokens
Why This Matters
Every AI Agent execution now:
Uses fewer tokens
Costs less to run
Maintains the same response quality
This allows:
More workflow executions within the same budget
Better scalability for AI-powered automation
Improved efficiency across all AI use cases
Common Issues and Fixes
Output looks shorter than before
This is expected due to removal of redundant data
Core response quality and accuracy are unchanged
Concern about missing context
Relevant context is still preserved
Older data is summarized, not removed
Token usage still high
Review workflow design (large inputs, long prompts, or many actions can still increase usage)
FAQ
Does this affect AI response quality?
No. Output quality remains the same. Only redundant or unnecessary data has been removed.
Do I need to enable this optimization?
No. This improvement is automatic for all AI Agent actions.
Are older conversations still used?
Yes. Older interactions are summarized while recent ones are kept in full detail.
Will this reduce my AI costs?
Yes. Since billing is based on tokens, reduced usage directly lowers cost.
Can I control how much context is sent?
Not directly. The system automatically determines the most efficient context size.