Skip to Content
GuidesImplementation Best Practices

Implementation Best Practices

Get the most out of Resly by following these proven best practices from successful customers.

Knowledge Base Content

Do

✓ Start with your top 20 most-asked inbound questions ✓ Write in clear, simple language customers understand ✓ Include specific examples and step-by-step instructions ✓ Keep articles focused on one topic each ✓ Update content when policies or products change

Don’t

✗ Upload technical jargon or internal documentation ✗ Include outdated or contradictory information ✗ Write 5000-word articles when 500 will do ✗ Forget to include common variations of questions ✗ Leave gaps in critical workflows

Quality over quantity: 10 well-written articles covering common questions outperform 100 poorly organized docs.


AI Tone and Guidance

Do

✓ Match your brand voice (friendly, professional, casual, empathetic) ✓ Give specific examples of how to respond to edge cases ✓ Set clear boundaries (what the AI should/shouldn’t do) ✓ Test different tones with real customer questions ✓ Iterate based on how customers respond

Don’t

✗ Use generic “be helpful” guidance ✗ Contradict your knowledge base content ✗ Tell the AI to do things it can’t (e.g., access accounts it can’t see) ✗ Change tone drastically after customers are used to it

Example good guidance:

“Always mention our 30-day guarantee when discussing returns. Never promise refunds over $500 without manager approval. Escalate billing questions about enterprise contracts.”


Widget Placement

Do

✓ Put chat widget in bottom-right (standard for most sites) ✓ Put search launcher in bottom-left if using both ✓ Test on mobile to ensure launchers don’t overlap content ✓ Make sure both are visible above fold ✓ Use contrasting colors so they’re easy to spot

Don’t

✗ Hide widgets behind other page elements ✗ Use accent colors that clash with your brand ✗ Position launchers where they cover critical content ✗ Forget to test on tablets and mobile devices


Testing Before Launch

Do

✓ Test with at least 20 real customer questions ✓ Review the event log to see what knowledge the AI retrieves ✓ Try edge cases and off-topic questions ✓ Have multiple team members test independently ✓ Check on different browsers and devices

Don’t

✗ Launch without testing in preview mode first ✗ Assume it’ll work perfectly on day one ✗ Skip testing on mobile ✗ Test only happy-path scenarios

Always test in preview mode before deploying to production. Verify the AI retrieves the right content and provides accurate answers.


Monitoring and Iteration

Do

✓ Check the Analyze tab weekly for common questions ✓ Look for patterns in what customers ask ✓ Add knowledge base content for frequently asked questions ✓ Monitor routed cases to see what the AI can’t handle ✓ Refine AI guidance based on real interactions

Don’t

✗ Set it and forget it ✗ Ignore customer feedback about AI responses ✗ Let knowledge base get stale ✗ Keep content that’s no longer accurate

Weekly review checklist:

  1. Check top 10 most-asked questions
  2. Identify content gaps (questions AI can’t answer)
  3. Review escalated cases for patterns
  4. Add/update 2-3 articles based on findings
  5. Test improvements in preview mode

Escalation Strategy

Do

✓ Set clear criteria for when to escalate to humans ✓ Make escalation easy (one-click case creation) ✓ Include conversation context in escalated cases ✓ Respond to escalated cases promptly ✓ Use case feedback to improve knowledge base

Don’t

✗ Make customers ask three times before escalating ✗ Lose conversation context during handoff ✗ Let escalated cases sit unread ✗ Escalate everything (defeats the purpose)

Good escalation triggers:

  • Customer explicitly requests human follow-up
  • AI confidence is low (<0.7 relevance)
  • Question is outside knowledge base
  • High-value transaction needs approval
  • Customer is frustrated or angry

Measuring Success

Key Metrics to Track

  1. Resolution Rate - % of questions answered without escalation
  2. Average Response Time - Should be <2 seconds
  3. Customer Satisfaction - From feedback buttons
  4. Time Saved - Hours not spent on repetitive questions
  5. Knowledge Base Coverage - % of questions AI can answer

Expected Performance

TimeframeResolution RateResponse TimeSatisfaction
First month40-60%<2s75%+
After 3 months70-85%<2s85%+
After 6 months80-94%<2s90%+

Performance improves over time as you add content and refine based on customer interactions.


Content Organization

Article Structure

Good article structure:

# Clear Title (matches how customers phrase it) Brief intro explaining what this article covers. ## Step 1: First Action Specific instructions with examples. ## Step 2: Next Action More details. ## Common Issues - Problem: Solution - Problem: Solution ## Need More Help? Link to related articles or escalation option.

Topic Coverage

Priority 1 (create first):

  • Top 20 inbound sales and service questions
  • Getting started guides
  • Common troubleshooting
  • Policy information (returns, shipping, etc.)

Priority 2 (add next):

  • Product/service details
  • Account management
  • Advanced features
  • Integration guides

Priority 3 (nice to have):

  • Best practices
  • Tips and tricks
  • Industry-specific guides
  • Edge cases

Continuous Improvement

The best Resly implementations improve over time:

Monthly Review Process

Week 1: Analyze performance

  • Review resolution rate trends
  • Check customer satisfaction scores
  • Identify top unanswered questions

Week 2: Plan improvements

  • Prioritize content gaps
  • Draft new articles
  • Update outdated content

Week 3: Implement changes

  • Add/update knowledge base
  • Refine AI guidance
  • Test in preview mode

Week 4: Monitor impact

  • Check if resolution rate improved
  • Review new escalation patterns
  • Adjust as needed

Consistency is key: Small, regular improvements outperform occasional large updates.


Common Pitfalls to Avoid

1. Insufficient Knowledge Base

Problem: Deploying with too little content Solution: Start with minimum 20 articles covering common questions

2. Unclear AI Guidance

Problem: Generic instructions like “be helpful” Solution: Provide specific examples and boundaries

3. No Testing

Problem: Deploying without thorough testing Solution: Test with 20+ real questions before launch

4. Set and Forget

Problem: Not monitoring or improving after launch Solution: Weekly reviews and continuous iteration

5. Wrong Tone

Problem: AI voice doesn’t match brand Solution: Test different tones, use specific guidance


Advanced Tips

For E-commerce

  • Upload return and shipping policies first
  • Configure order lookup integration early
  • Set maximum refund amounts for safety
  • Add product FAQs by category

For SaaS

  • Start with onboarding documentation
  • Add API reference for developers
  • Create troubleshooting guides by feature
  • Include billing and account management

For Documentation Sites

  • Organize content by topic hierarchy
  • Use instant search mode for speed
  • Include code examples in articles
  • Link related documentation

Next Steps

Remember: The best implementations start simple and improve continuously based on real customer interactions.

Last updated on