AI Workflow & Governance Lab
Build AI controls around business risk, not model hype.
This lab focuses on the decisions that separate a useful enterprise AI workflow from a chatbot demo: grounding, risk, approval, auditability and behavioural testing.
Model output ≠ business authority
Controls follow impact and risk
Test behaviour, not fluency
Confidence is high. Is approval still needed?
An AI assistant drafts a refund response. It reports 94% confidence, but sending the message would commit the business to a $1,200 refund.
Grounding before generation
A customer asks whether an event ticket can be transferred. The model knows general ticketing practices, but the organiser has a specific transfer policy.
What belongs in the audit log?
An AI-generated reply is edited by a staff member before it is sent.
Testing an AI workflow
A team says the feature is tested because the API returns 200 and the model produces fluent text.
Your result
0 / 4
Complete all four scenarios to finish the lab.