All learning modules

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.

Learn by making an engineering decision, then inspect the reasoning.

Model output ≠ business authority

Controls follow impact and risk

🧪

Test behaviour, not fluency

1

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.

2

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.

3

What belongs in the audit log?

An AI-generated reply is edited by a staff member before it is sent.

4

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.