Learning Timeline
Key Insights
The Importance of Dynamic Timestamps
AI models like Claude or GPT have a 'knowledge cutoff'. If you don't explicitly provide today's date, the AI might set task dates for the wrong year (e.g., 2021 instead of 2024).
Google Calendar Field Naming Tips
In the n8n Google Calendar tool, the field for the event name is called 'Summary'. Don't get confused looking for fields named 'Title' or 'Event Name' as they do not exist by default.
The Advantage of 'Let Model Decide'
Using this feature is much more flexible than manual mapping. The AI can intelligently extract relevant information even if the sentence structure in the voice note varies.
Prompts
Dynamic Timestamp Injection
Target:
n8n System Prompt
The current date and time is: {{ $now }}
Step by Step
Configuring Agent Tools with 'Let Model Decide' Parameters
- Open the AI Agent node in your n8n canvas.
- Identify the tool you want to configure (e.g., AirTable, Google Calendar, or Notion).
- For each input field (such as Name, Description, or Title), click the 'sparkly' icon (small star) to the right of the input field.
- Select 'Let the model decide' so the AI determines the field content based on the context from the transcript.
- In the Google Calendar node, click 'Add Field' and select 'Summary' to set the event title.
- Set both the 'Summary' and 'Description' fields to 'Let the model decide'.
- In the Notion node, ensure the 'Title' and 'Date' fields are also set to 'Let the model decide'.
- Specify the Timezone in the Notion node if you are using date fields to ensure timing accuracy.
Adding Real-Time Context (Timestamp) to the Agent Prompt
- Open the 'System Message' or 'Prompt' section within the AI Agent node.
- Add a line at the bottom of the prompt: 'The current date and time is:'.
- Change the prompt input mode to 'Expression'.
- Drag or insert the timestamp variable (from the previous code node) into the prompt after that sentence.
- This step is crucial to prevent the AI from using outdated dates from its training data.
Running End-to-End Workflow Testing
- Click the 'Execute Workflow' button in n8n to put the workflow in standby mode.
- Send a voice note via Slack containing complex instructions (e.g., Log a content idea to AirTable AND create a task in Notion).
- Monitor the n8n panel to see the AI Agent start processing the instructions.
- Check the execution logs; the agent might attempt to call the tool multiple times if there is a formatting error on the first try (self-correction).
- Open Slack to see the feedback message sent by the bot.
- Click the Notion link or check AirTable to ensure the data was accurately entered according to your voice instruction.
