Impact assessments missing for two customer-facing AI systems
What we found
The organization had an AI risk register, but no documented impact assessment for its customer support chatbot or its credit pre-screening model. Neither assessed the consequences for the individuals affected.
Impact
Likely nonconformity at the certification audit, and no evidence for customers who ask how affected individuals are protected.
How to fix it
Run and document impact assessments for both systems, link the results to risk treatment, and set a trigger to repeat them when a system changes.






