Hi Lindy team,
I'm building an agent that retrieves a specific Google Spreadsheet cell and posts the value to a Slack channel both on a weekly schedule and on-demand. The scheduled trigger works well, but on-demand triggering required configuring a keyword-based Slack trigger, and even then the agent only works when I manually add a custom step prompt inside the workflow.
The current issues are:
- 1.
The response is very slow — often 30–60 seconds, sometimes several minutes.
- 2.
Some model choices fail entirely (e.g., GPT-4.1, GPT-5.1-fast), while Haiku works but slowly.
Could you advise how to optimize the agent for sub-second or near-instant (<1s) response time for simple spreadsheet-to-Slack retrieval? Any best practices for model selection, step design, or caching would be greatly appreciated.
Thank you!