Auto processing - Lancey Docs

Overview

Auto Processing enables you to define custom prompts that tell Lancey which tasks it should automatically work on. Instead of manually triggering tasks, Lancey monitors incoming feedback and customer issues, then intelligently decides which ones match your criteria to start processing automatically.

Auto Processing is perfect for teams that want to reduce manual task creation while maintaining control over which issues Lancey should tackle. Set it once and let Lancey handle the rest.

How it works

Auto Processing follows a simple workflow:

You define a prompt - Write clear instructions about what types of tasks Lancey should work on
Lancey monitors feedback - Your integrations (Slack, Intercom, GitHub issues) feed customer feedback to Lancey
AI evaluation - Lancey evaluates each piece of feedback against your custom prompt
Automatic execution - Matching tasks are automatically added to your queue and processing begins
Team review - Your team can review and approve tasks before they’re merged

Setup

Navigate to Auto Processing

Go to Settings —> Issues in your Lancey dashboard

Add your processing instructions

In the Processing instructions text field, provide specific instructions for how the AI should evaluate and process issues

Write clear instructions

Describe the types of tasks Lancey should automatically process. Be specific about:

Issue types (bugs, features, enhancements)
Priority levels
Source integrations (Slack, Intercom, GitHub)
Any specific keywords or patterns

Enable Auto Processing

Toggle the Auto Processing switch to enable automatic processing for your rules

Save your configuration

Click Save to apply your processing instructions. Lancey will now automatically process matching issues based on your rules

Writing effective prompts

The quality of your auto-processing depends on how well you write your prompts. Here are key principles:

Be specific

Use concrete criteria rather than vague descriptions

Use examples

Include examples of what should and shouldn’t be processed

Define priority

Specify minimum severity/priority levels to process

Set boundaries

Clearly state what NOT to process to avoid false positives

Prompt examples

Example 1: Customer bug fixes

Process any GitHub issues or customer reports that describe:
- Bugs affecting production or user experience
- Error messages from users
- Features that don't work as documented

Do NOT process:
- Feature requests or enhancement suggestions
- Questions or documentation clarifications
- Issues related to non-core functionality

Priority: Only process issues marked as "critical" or "high" priority
Source: GitHub issues and Intercom support conversations

Example 2: High-priority enhancements

Automatically work on enhancement requests that:
- Come from paying customers
- Have been upvoted or requested by multiple users
- Are tagged with "high-value-customer" or "revenue-impacting"
- Relate to core product features

Skip:
- Low-priority nice-to-haves
- Requests from free tier users
- Niche or edge-case features

Severity: Only process if marked as "medium" priority or higher
Source: Intercom feature requests and GitHub discussions

Example 3: Security and performance fixes

Process all tasks related to:
- Security vulnerabilities or potential exploits
- Performance issues (slow endpoints, memory leaks)
- Data integrity problems
- Dependency updates and patching

Exclude:
- Cosmetic issues
- Minor UI improvements
- Non-breaking technical debt

Timeline: These tasks should be processed with highest urgency
Source: GitHub issues, security alerts, and internal reports

Example 4: API and integration issues

Automatically handle tasks for:
- External API integration failures
- Third-party service connectivity problems
- Webhook or callback issues
- Deprecated API endpoint migrations

Do NOT process:
- Internal service connectivity issues (handle manually)
- Issues requiring database schema changes
- Tasks needing manual data migration

Severity: Process if there's active customer impact
Source: Slack alerts and GitHub integration issues

Example 5: Documentation and typo fixes

Process low-effort improvements:
- Typos in documentation or code comments
- Broken links in docs or README files
- Outdated code examples
- Missing API documentation

Exclude:
- Major documentation rewrites
- Complex API documentation from scratch
- Content requiring SME review

Batch size: Process up to 5 of these tasks daily
Source: GitHub issues labeled "documentation" and "typo"

Best practices

Start simple

Begin with a narrowly focused prompt that targets a specific type of task. Once you see it working well, you can expand or create additional prompts for other use cases.

Monitor accuracy

Review which tasks Lancey auto-processes during the first week. Refine your prompt if you see too many false positives or misses important issues.

Use keywords strategically

Include specific keywords, labels, and patterns from your integrations. Reference your Slack channels, GitHub labels, and Intercom tags explicitly.

Set realistic expectations

Prompt quality affects results. A vague prompt like “fix bugs” will have lower accuracy than “fix critical bugs affecting the login flow reported in Slack #bugs channel.”

Review before merge

Even with auto-processing enabled, always require manual review before tasks are merged. This maintains code quality and prevents unintended changes.

Iterate based on results

Track which auto-processed tasks succeed or fail. Use this data to refine your prompt over time.

Troubleshooting

Lancey isn't processing any tasks?

Check that:

Your prompt is enabled in the Auto Processing settings
Your integrations (Slack, Intercom, GitHub) are properly configured
Your feedback sources are actively sending messages
Your prompt criteria match the incoming feedback

Too many false positives?

Your prompt might be too broad. Try:

Adding more specific keywords or criteria
Increasing priority thresholds
Excluding more edge cases
Testing with fewer sources initially

Tasks aren't being processed fast enough?

Verify that:

Your AI model configuration is set up correctly
Your GitHub repository and credentials are valid
Your team has available capacity to review PRs
Your prompt isn’t requiring unnecessary manual approval steps

How do I disable or modify a prompt?

You can:

Temporarily disable a prompt by toggling it off in Auto Processing settings
Edit an existing prompt to adjust criteria
Delete a prompt and create a new one
Create multiple prompts for different task types

Next steps

Create Tasks

Learn how to manually create and manage tasks

Modify Tasks

Update and refine your tasks

Integrations

Configure integrations to feed feedback to Lancey

Need help crafting the perfect auto-processing prompt? Reach out via the livechat widget on our platform for quick support or schedule a call with our team.

​Overview

​How it works

​Setup

​Writing effective prompts

Be specific

Use examples

Define priority

Set boundaries

​Prompt examples

​Example 1: Customer bug fixes

​Example 2: High-priority enhancements

​Example 3: Security and performance fixes

​Example 4: API and integration issues

​Example 5: Documentation and typo fixes

​Best practices

​Troubleshooting

​Next steps

Create Tasks

Modify Tasks

Integrations

Overview

How it works

Setup

Writing effective prompts

Prompt examples

Example 1: Customer bug fixes

Example 2: High-priority enhancements

Example 3: Security and performance fixes

Example 4: API and integration issues

Example 5: Documentation and typo fixes

Best practices

Troubleshooting

Next steps