Course Format and Delivery

Delivery Method: Instructor-led LiveOnline

About this course

Generative AI is already in the workflow: research, synthesis, first drafts and client-facing deliverables. The risk is no longer that professionals fail to adopt the tools. The risk is that they use them well enough to stop checking them, and that agentic systems arrive before the habit of checking is established. In a peer-reviewed survey of 319 knowledge workers reporting 936 real uses of generative AI, higher confidence in the AI predicted measurably less critical thinking. Higher confidence in one's own expertise predicted more.

This course is built for practitioners who already know how to prompt and now need to know what to trust. It covers how these systems work and where they predictably fail, how to verify proportionately to risk, how to diagnose and correct a failing output rather than re-rolling it, and what changes when work moves to an agent that plans, calls tools and acts. Two gates, data quality and regulatory exposure, decide whether work should touch an agent at all.

What you will learn

• Explain how large language models generate output, and predict where they will fail

• Recognize the five failure modes of generative AI in live output

• Verify proportionately to the consequence of a deliverable being wrong

• Diagnose a failing output and apply the right correction rather than re-prompting

• Judge whether data is fit for agentic use, assisted use, or neither

• Identify the binding constraint across regulation, client policy and firm policy

• Decompose a workstream and assign each step to human, GenAI or agentic execution

• Place oversight gates where they catch premises rather than finished artifacts

• Structure accountability so that automation bias is reduced rather than assumed away

This course is great for

Active, experienced consultants who produce client-facing analysis with AI assistance, Senior Associates through Managing Directors, plus risk, quality and practice leadership responsible for the firm's AI posture.

Topics Covered

1. Frame and Baseline. The three shifts: gathering to verifying, solving to integrating, doing to supervising. Why confidence in the tool predicts less critical thinking. Exercise: The Unchecked Line.

2. How These Systems Work, and Where They Break. Prediction, not retrieval. Five failure modes. Verification proportionate to consequence. Exercise: Confident and Wrong.

3. Critical Thinking with GenAI: Verify, Then Correct. Four bucket triage. Correction as diagnosis, not retry. Accountability as a control. Exercise: Triage and Recovery Drill.

4. Critical Thinking with AI in Professional Services. Using Claude across the consulting lifecycle, from problem framing to client delivery. Team case: The AI-Assisted Consulting Challenge.

5. Data Quality as the First Gate. Four dimensions of quality, fitness for use, the three-way verdict. Exercise: Fitness-for-Use Audit.

6. Regulation, Client Policy and the Second Gate. Finding the binding constraint before the work starts. What is genuinely unsettled. Exercise: The Binding Constraint.

7. Systems Thinking: Designing How You Attack the Work. Decompose a real workstream and assign each step to human, GenAI or agent. Exercise: Attack Design.

8. Building Your Personal Standard. The arc, and one change on a named engagement within two weeks. Exercise: The Commitment.