Customer Support Chatbot

Limited Risk

LLM-powered virtual assistant handling Tier-1 customer inquiries. Operates 24/7 across web and mobile channels with human escalation.

91/ 100

Assessment ProgressComplete

EU AI ActISO 42001

Assessment Questionnaire

Transparency

Are users clearly informed they are interacting with an AI chatbot, not a human?

Yes — prominent disclosure at conversation start and in the UI header. Compliant with EU AI Act Art. 52 transparency obligation for AI systems interacting with natural persons.

EU AI Act

Does the chatbot disclose its limitations and suggest human escalation when uncertain?

Yes — confidence thresholds trigger automatic escalation message. Below 0.6 confidence, users are directed to a human agent.

ISO 42001

Data Governance

Is personal data collected during chat sessions handled in accordance with privacy requirements?

Yes — PII is redacted from training logs. Conversation data retained for 90 days for quality assurance, then anonymized.

ISO 42001

Content Safety

Are guardrails in place to prevent the chatbot from generating harmful, misleading, or biased content?

Yes — content filtering pipeline with toxicity detection, hallucination checks, and brand-safety rules. Outputs are post-processed before delivery.

EU AI Act

Human Oversight

Is there a mechanism for human agents to review and correct chatbot responses?

Yes — flagged conversations are routed to human review queue. Corrections feed back into fine-tuning pipeline.

EU AI Act

Accuracy

Are chatbot response accuracy and customer satisfaction metrics monitored?

Resolution rate: 82%, CSAT: 4.2/5, hallucination rate: 1.3% — monitored daily with weekly review meetings.

ISO 42001

Record Keeping

Are conversation logs maintained with appropriate access controls?

Yes — all conversations logged with role-based access. Retention: 90 days full, 2 years anonymized aggregates.

ISO 42001

Gap Analysis

1 gap identified

A.8.4 — AI System Operationmedium

Escalation threshold (0.6 confidence) not validated against recent production data.

Recommendation: Run calibration study on last 3 months of conversations to validate or adjust escalation threshold.

Back to Assessments

Customer Support Chatbot

Limited Risk

LLM-powered virtual assistant handling Tier-1 customer inquiries. Operates 24/7 across web and mobile channels with human escalation.

91/ 100

Assessment ProgressComplete

EU AI ActISO 42001

Assessment Questionnaire

Transparency

Are users clearly informed they are interacting with an AI chatbot, not a human?

Yes — prominent disclosure at conversation start and in the UI header. Compliant with EU AI Act Art. 52 transparency obligation for AI systems interacting with natural persons.

EU AI Act

Does the chatbot disclose its limitations and suggest human escalation when uncertain?

Yes — confidence thresholds trigger automatic escalation message. Below 0.6 confidence, users are directed to a human agent.

ISO 42001

Data Governance

Is personal data collected during chat sessions handled in accordance with privacy requirements?

Yes — PII is redacted from training logs. Conversation data retained for 90 days for quality assurance, then anonymized.

ISO 42001

Content Safety

Are guardrails in place to prevent the chatbot from generating harmful, misleading, or biased content?

Yes — content filtering pipeline with toxicity detection, hallucination checks, and brand-safety rules. Outputs are post-processed before delivery.

EU AI Act

Human Oversight

Is there a mechanism for human agents to review and correct chatbot responses?

Yes — flagged conversations are routed to human review queue. Corrections feed back into fine-tuning pipeline.

EU AI Act

Accuracy

Are chatbot response accuracy and customer satisfaction metrics monitored?

Resolution rate: 82%, CSAT: 4.2/5, hallucination rate: 1.3% — monitored daily with weekly review meetings.

ISO 42001

Record Keeping

Are conversation logs maintained with appropriate access controls?

Yes — all conversations logged with role-based access. Retention: 90 days full, 2 years anonymized aggregates.

ISO 42001

Gap Analysis

1 gap identified

A.8.4 — AI System Operationmedium

Escalation threshold (0.6 confidence) not validated against recent production data.

Recommendation: Run calibration study on last 3 months of conversations to validate or adjust escalation threshold.