Skip to main content

AI Best Practices

Get the most out of Gover's AI Assistant with these tips.

Writing Good Prompts

Be Specific

❌ "Help me with compliance" ✅ "What controls should I implement for GDPR Article 32 security requirements?"

Provide Context

❌ "Suggest mappings" ✅ "Suggest controls for the ISO 27001 A.9 Access Control requirements"

Ask Follow-ups

Build on previous responses:

  • "Can you explain that in more detail?"
  • "What about for smaller organizations?"
  • "Show me examples"

Working with Suggestions

Start with High Confidence

Accept suggestions with 80%+ confidence first, then review lower scores.

Verify Before Accepting

Always review suggestions against your actual controls and requirements.

Use Bulk Actions Wisely

Bulk accept only when you've reviewed the suggestions.

Provide Feedback

Accepting and rejecting suggestions helps AI learn your preferences.

Improving AI Quality

Better Descriptions

Detailed control and requirement descriptions lead to better suggestions.

Consistent Naming

Use consistent terminology across your workspace.

More Data

AI improves as you add more controls, documents, and mappings.

Regular Use

The more you use AI features, the better they become.

When to Use AI

Good Use Cases

  • Initial mapping of new frameworks
  • Finding gaps in coverage
  • Drafting descriptions
  • Exploring compliance questions

When to Be Careful

  • Final audit preparation (always verify)
  • Highly specialized requirements
  • Organization-specific policies

Limitations

AI is helpful but has limitations:

LimitationMitigation
May miss contextProvide more details
Can be wrongAlways verify suggestions
Doesn't know your orgReview against your reality
Generic adviceCustomize for your needs

Privacy Reminders

  • AI doesn't store your prompts
  • Your data stays in your workspace
  • No training on your data
  • EU data residency (Sweden Central)

Summary

  1. Be specific in your questions
  2. Provide context about what you're working on
  3. Review suggestions before accepting
  4. Verify against reality — AI isn't perfect
  5. Use regularly to improve quality

Next Steps