Troubleshooting & Support
Find solutions to common problems and get your AI Product Manager working smoothly.
Content Quality Issues
Generated Content is Too Generic
Symptoms:
Output lacks specific details, uses standard language without customization
Solutions:
- →Provide more detailed input in product description
- →Include specific features, unique selling points, and differentiators
- →Add examples of similar products or features you want to emulate
- →Regenerate with additional constraints and requirements
- →Try a different AI model (Claude typically produces less generic output)
Missing Important Features or Sections
Symptoms:
PRD skips key requirements, architecture misses components
Solutions:
- →Check if you mentioned all features in input
- →Regenerate with more comprehensive product description
- →Manually add missing sections (system allows full editing)
- →Use PRD as draft and refine with team expertise
- →Try with a different model or configuration
Incorrect Architecture Recommendations
Symptoms:
Tech stack doesn't match our preferences, missing integration points
Solutions:
- →Explicitly state technology preferences in product description
- →Mention existing systems and integrations
- →Specify performance requirements and scalability needs
- →Manually edit architecture section with correct recommendations
- →Share company tech standards as input for regeneration
Tasks Are Too Large or Too Small
Symptoms:
Generated tasks don't match team's estimation practices
Solutions:
- →Manually adjust story points based on team experience
- →Break large tasks into smaller subtasks
- →Combine very small tasks into single tickets
- →Regenerate tasks with specific scope guidance
- →Establish team standards then customize output
Timeline & Estimation Issues
Timeline Estimates Are Unrealistic
Symptoms:
Timeline is too optimistic or too pessimistic for your team
Solutions:
- →Have engineering lead review all estimates
- →Adjust for team experience (junior teams need more time)
- →Account for team size and allocation
- →Factor in technology familiarity
- →Add 15-20% buffer for unknowns
- →Use historical project data to calibrate
Phases Are Too Aggressive
Symptoms:
Sprint durations seem too short, dependencies not properly mapped
Solutions:
- →Extend individual sprint durations
- →Re-sequence phases to reduce dependencies
- →Break phases into smaller increments
- →Regenerate roadmap with specific duration constraints
- →Get team input on realistic sprint pacing
Missing Critical Path Issues
Symptoms:
Roadmap doesn't account for hard dependencies
Solutions:
- →Manually identify and mark critical dependencies
- →Regenerate with explicit dependency information
- →Create dependency diagram to share with AI for regeneration
- →Have engineering team review and correct dependency order
Technical Issues
Generation Timeout or Takes Too Long
Symptoms:
Generation appears stuck, takes 10+ minutes to complete
Solutions:
- →Try switching to faster model (Groq is 50% faster)
- →Reduce scope (generate PRD only instead of full planning)
- →Simplify product description to reduce complexity
- →Try again in off-peak hours (avoids server congestion)
- →Contact support if consistent timeouts occur
Export to Project Management Tool Fails
Symptoms:
Integration sync doesn't work, tasks don't appear in Jira/Linear/Asana
Solutions:
- →Verify API connection is properly authenticated
- →Check that you have write permissions in target tool
- →Ensure project/board is selected before sync
- →Try exporting as CSV and manual import instead
- →Contact support with error message
Upload File Fails or Shows Invalid Format
Symptoms:
Can't upload existing document, unsupported format error
Solutions:
- →Check file size is under 10MB
- →Verify file format is supported (PDF, Word, Markdown, TXT, CSV)
- →Convert file to supported format if needed
- →Try removing images/attachments from original file
- →Copy-paste content instead of uploading
Diagrams or Formatting Look Broken
Symptoms:
Mermaid diagrams don't render, formatting is corrupted
Solutions:
- →Try regenerating just the architecture section
- →Switch export format (PDF, Markdown, JSON)
- →Copy raw Mermaid code and render in separate tool
- →Manually recreate diagram in preferred tool
- →Report to support with screenshot
Workflow Issues
Can't Find Previous Version
Symptoms:
Version history is empty, can't revert to earlier state
Solutions:
- →Check version history panel (should show all generations)
- →Try refreshing page to reload version list
- →Filter by date range to find specific version
- →Contact support if versions appear to be deleted
- →Always create snapshots before major changes
Team Member Can't See Shared Document
Symptoms:
User doesn't have access despite being invited
Solutions:
- →Check user has accepted invitation
- →Verify user has correct permission level (view vs edit)
- →Resend invitation link to user
- →Add user directly from access settings
- →Try logging out and back in
Changes Not Saving
Symptoms:
Edits to document don't persist after refresh
Solutions:
- →Check internet connection is stable
- →Look for save indicator (should show saved status)
- →Try refreshing page to see if changes persisted
- →Clear browser cache and try again
- →Switch browsers to test if issue is browser-specific
AI Model & Quality Issues
Claude Generation Fails Midway
Symptoms:
Generation starts but stops partway through
Solutions:
- →Try with Groq or OpenRouter as alternative
- →Reduce scope of generation (fewer requirements)
- →Simplify product description
- →Wait a few minutes and try again
- →Contact support if consistently fails
Different Models Give Very Different Results
Symptoms:
Outputs from Claude vs Groq are significantly different
Solutions:
- →This is expected - different models have different strengths
- →Claude better for nuance and quality
- →Groq better for speed and consistency
- →Use Claude for important/complex projects
- →Use Groq for quick drafts and iteration
Model Keeps Generating Same Mistakes
Symptoms:
Regenerating with same model doesn't fix issues
Solutions:
- →Modify your input description to address the issue
- →Be very specific about what you want differently
- →Try different model for different output
- →Make partial edit then regenerate that section
- →Create different project with revised input