Skip to main content
Best practices for deploying and using Kafka effectively in your organization.

​
Start Small, Scale Gradually

Begin with one use case Don’t try to automate everything at once. Pick one high-value, repetitive task and build a single AI Employee around it. Good first use cases:
  • Daily email summaries
  • Candidate pipeline updates
  • Task syncing between tools
Expand after validation Once your first AI Employee proves valuable, add more workflows and playbooks. Then consider creating additional AI Employees for different roles.

​
Define Clear Boundaries

Set explicit escalation rules In Global Instructions, specify exactly when Kafka should ask for human input vs. proceeding autonomously. 
Escalate to me if:
- Financial decisions over $500
- Candidate has competing offers
- Project timeline changes by more than 1 week
Use activation criteria thoughtfully Make playbook triggers specific enough to activate when needed, but not so broad they fire constantly.

​
Iterate Based on Real Use

Deploy, observe, refine Treat your AI Employee as a living system:
  1. Deploy in real work scenarios
  2. Monitor how Kafka performs
  3. Adjust Global Instructions, Workflows, and Playbooks
  4. Repeat weekly
Track what works Keep notes on:
  • Which workflows save the most time
  • Which playbooks get used most often
  • Where Kafka struggles or needs clarification

​
Give Kafka Context

Be specific in instructions Bad: “Check my calendar” Good: “Check my @google-calendar for conflicts between 2-4pm tomorrow” Bad: “Find candidates” Good: “Use @apollo to find 10 software engineers in San Francisco with 5+ years React experience” Provide examples In Global Instructions and Playbooks, show Kafka what good output looks like. 
When summarizing emails, format like this:
• [Sender] - [Topic] - [Action needed]
• John Smith - Q4 Budget - Needs approval by Friday

​
Use the Right Tool for the Job

Workflows → Fully autonomous, recurring tasks
  • Daily summaries
  • Scheduled reports
  • Event-driven notifications
Playbooks → Semi-structured tasks needing consistency
  • Standard procedures
  • Quality-controlled processes
  • Repeatable formats
Direct conversation → One-off requests
  • Ad-hoc questions
  • Non-repeatable tasks
  • Exploratory work

​
Enable Team Adoption

Add Kafka to relevant Slack channels Put Kafka where the work happens. If your team discusses recruiting in #hiring, add your Recruiter AI Employee there. Document your AI Employees Keep a simple doc explaining:
  • What each AI Employee does
  • How to interact with it
  • When to use workflows vs. asking directly
  • Escalation protocols
Gather team feedback Your team will discover edge cases and improvement opportunities. Create a channel for feedback.

​
Security & Privacy

Grant minimum necessary permissions When connecting integrations, give Kafka only the access it needs for its specific role. Review sensitive data handling In Global Instructions, specify how Kafka should handle:
  • Confidential information
  • Personal data
  • Financial information
  • Customer data
Use private channels for sensitive work For AI Employees handling confidential information, use private Slack channels with restricted access.

​
Measure Impact

Track these metrics:
  • Time saved — Hours per week your team gains back
  • Tasks automated — Number of manual tasks Kafka handles
  • Response time — How quickly Kafka completes requests
  • Error rate — How often Kafka needs correction
Adjust your approach based on what delivers the most value.

​
Common Pitfalls to Avoid

Over-automating too quickly Don’t automate complex processes before validating simple ones. Vague instructions “Handle my email” won’t work well. “Scan @gmail daily at 9am for emails from clients and summarize urgent ones” will. Ignoring feedback loops If Kafka makes the same mistake repeatedly, update Global Instructions or the relevant Playbook. Forgetting to update integrations As your stack changes, keep Kafka’s connected tools current.

​
Getting Help

We’re actively supporting early customers. Email: [email protected] In-app support: Use the support widget for quick questions Share your use cases and challenges — your feedback shapes the product.