Exercise 1: The Task Audit

What you're doing: Identifying every recurring task in your role and categorizing it by what AI can and can't do.

Time estimate: 45–60 minutes

Instructions:

Open a document. Write down every task you do regularly — daily, weekly, or monthly. Don't filter. Include the small ones. Update the CRM. Write reports. Lead meetings. Give feedback. Approve expenses. Review code. Respond to email. Unblock team members. Everything.

Aim for at least fifteen tasks. If you can only name ten, you're thinking too abstractly. Get specific.

Now categorize each task into one of three columns:

Column A: AI can do this now. These are tasks where AI has capabilities that would let someone without your skills execute the task, with minimal direction. Writing emails from a brief. Summarizing documents. Generating code. Analyzing datasets. Creating slide decks. Running basic research. These tasks are not gone, but the skill barrier has changed.

Column B: AI can assist, but human judgment required. These are tasks where AI can handle significant portions — drafting, initial analysis, generating options — but the decision-making, shaping, and quality judgment still require you. Editing AI-written copy for voice and strategy. Reviewing AI-generated code for architecture and fit. Evaluating AI analysis for what it means in context. Deciding what customer feedback to implement.

Column C: AI cannot do this reliably. These are tasks that depend on judgment, context, relationships, or decisions that are hard to codify. Knowing which customer problem actually matters. Deciding what your team should focus on next quarter. Recognizing when someone is struggling and how to help. Building trust with a new executive stakeholder. Navigating a difficult conversation. These are not necessarily harder tasks. They're tasks that depend on things AI can't supply: relationships, judgment born from experience, the ability to read a room.

What to notice:

How much of your time lives in each column? If most of it is in Column A, that's useful information. If it's split across all three, that's also useful. There's no right distribution. The point is to see it clearly.

Are there tasks in Column A that you thought were Column C? That's the denial mode talking. Be honest.

Are there tasks in Column C that might actually belong in Column B? That's probably the panic mode. Not everything you do is irreplaceable. But not everything is automatable either.

What you're building toward: understanding what percentage of your role is execution (Columns A and B) versus judgment and direction (Column C).