Your AI agent answers, decides, and edits your files without you: did you give it a framework?
Following up with a customer and giving them a discount, paying a supplier’s invoices, answering your emails, changing or deleting files: each of these actions is a power handed to an AI agent. This module gives you the method to grant that power without signing a blank check: a clear mandate, proportionate to the risk, that you can revise and answer for.
What you’ll take away
- the difference between giving an agent a task and giving it the power to act
- six clauses for a safe, useful mandate, from thresholds to stop points
- three questions to adjust the agent’s autonomy and know when to withdraw it
Who it’s for
Team members who want to give an AI agent real power, and managers who want to set the framework for it.
Learning objectives
- Distinguish delegating a task from delegating the power to act, and name the four powers given to agents: money, communication, systems, and people.
- Build a six-clause mandate: outcome, playing field, allowed and forbidden actions, thresholds, stop points, duration and record.
- Match the agent’s autonomy to the cost and reversibility of an error, and make forbidden actions impossible through configuration.
- Revise or withdraw a mandate using the activity log, and explain the accountability of the organization, manager, and team member.
How does this module work?
The steps of the Adaptive module.
- Self-assessmentOn nine dimensions, to gauge current habits and guide the practice.
- Expert videoA ~10 min video with three pauses for reflection.
- Practice scenariosThe first two scenarios chosen from the self-assessment, then sixteen more, optional. Every answer gets feedback: what it leads to, and one way to improve.
- Dated commitmentOne action to try, written by the learner or chosen from the suggestions.
- Memo to downloadKey takeaways, questions, commitment, personal notes, and checklists.
Duration: ~20 min. Excluding bonus practice scenarios and optional in-depth content.



