Pricing
    About UsCareersContactPartnersPressStatus
    Start FreeLog in
    AI System Inventory
    10 min readUpdated 2025-01-28

    Using the AI System Wizard

    Master the 20-step AI System Intake Wizard to capture comprehensive compliance information.

    The AI System Wizard

    The AI System Wizard is Klarvo's guided intake process, designed to capture all information needed for EU AI Act compliance in a structured, user-friendly flow.

    Wizard Design Philosophy

    The wizard follows these principles:

  1. < 10 minutes for a complete Full Assessment
  2. Every answer creates action: Classification decision, obligation flag, evidence request, or task
  3. Progressive disclosure: Complex questions only appear when relevant
  4. Save progress: Resume anytime without losing work

    5. The 20 Steps

    #### Part 1: Foundation (Steps 0-3)

    Step 0: Mode Selection

    Choose Quick Capture or Full Assessment. Quick Capture skips optional questions and creates follow-up tasks.

    Step 1: Basic Information

  5. System name and internal reference ID
  6. Lifecycle status (Idea, Pilot, Live, Retired)
  7. Primary owner assignment

    8. Step 2: Vendor Information

  8. Built in-house vs. third-party acquisition
  9. Vendor selection or creation
  10. Contract documentation links

    11. Step 3: Ownership & Accountability

  11. Backup owner
  12. Oversight owner
  13. Department/team assignment

    14. #### Part 2: Scope & Context (Steps 4-5)

    Step 4: Deployment Scope

  14. Deployment regions (EU, UK, US, Other)
  15. EU countries of operation
  16. Internal user groups
  17. Affected external parties

    18. Step 5: Value Chain Role

  18. Deployer, Provider, or Both
  19. External offering assessment
  20. Foundation model usage

    21. #### Part 3: Definition & Classification (Steps 6-9)

    Step 6: AI System Definition Test

  21. Infers outputs from inputs?
  22. Output types (predictions, recommendations, etc.)
  23. Operates autonomously?
  24. Adapts after deployment?
  25. Technical approach (ML, LLM, rules-based)

    26. Step 7: Use Case & Functionality

  26. Purpose category
  27. Workflow description
  28. Human involvement level
  29. Override capability
  30. Usage frequency and scale

    31. Step 8: Prohibited Practices Screening

    Eight critical questions covering Article 5:

  31. Manipulation/deception
  32. Exploitation of vulnerabilities
  33. Social scoring
  34. Criminal profiling
  35. Facial recognition scraping
  36. Workplace emotion inference
  37. Biometric categorisation
  38. Real-time remote biometric ID

    39. Step 9: High-Risk Screening

    Nine Annex III category checks:

  39. Biometrics
  40. Critical infrastructure
  41. Education
  42. Employment
  43. Essential services
  44. Law enforcement
  45. Migration/border
  46. Justice/democratic processes
  47. Safety components

    48. #### Part 4: Obligations (Steps 10-14)

    Step 10: Transparency Obligations

    Article 50 requirements:

  48. Direct interaction disclosure
  49. Synthetic content marking
  50. Emotion recognition notice
  51. Deepfake disclosure
  52. Public-interest text disclosure

    53. Step 11: Data & Privacy

  53. Personal data processing
  54. Special category data
  55. Minors involved
  56. Data sources and control
  57. Retention periods
  58. DPIA status

    59. Step 12: Human Oversight

  59. Oversight model (HITL/HOTL/HOOTL)
  60. Oversight owner authority
  61. Competence requirements
  62. Training status
  63. SOP documentation

    64. Step 13: Logging & Records

  64. Automatic logging capability
  65. Log storage location
  66. Access controls
  67. Retention period
  68. Export capability
  69. 6-month retention confirmation

    70. Step 14: Incidents & Monitoring

  70. Incident response process
  71. Severity levels
  72. Notification procedures
  73. Suspension capability

    74. #### Part 5: Special Contexts (Steps 15-17)

    Step 15: Workplace Context

  74. Workplace use assessment
  75. Worker notification status

    76. Step 16: Public Authority Context

  76. Public authority status
  77. Public service provision
  78. Registration status

    79. Step 17: Training & AI Literacy

  79. Staff roles involved
  80. Training program status
  81. Completion tracking

    82. #### Part 6: Finalization (Steps 18-20)

    Step 18: FRIA Assessment

  82. FRIA trigger evaluation
  83. FRIA status tracking
  84. Link to FRIA Wizard

    85. Step 19: Review & Sign-off

  85. Final classification confirmation
  86. Reviewer assignment
  87. Sign-off date
  88. Notes

    89. Step 20: Completion

  89. Summary of outputs
  90. Generated documents
  91. Next steps
  92. Action items

    93. Auto-Generated Outputs

    Upon completion, the wizard creates:

  93. AI System Record: Full database entry
  94. Classification Memo: PDF with rationale
  95. Gap Checklist: Missing controls/evidence
  96. Task Plan: Auto-assigned action items
  97. Evidence Requests: Specific document needs