Developing Agentic Leadership: Sustaining School Improvement with AI

School leaders must sustain improvement work alongside competing demands on their time and attention. Maintaining plans, recovering decisions and following up commitments can become a substantial task in itself. How might working with an AI agent help leaders sustain this work?

Drawing on my developing collaboration with Johnson, my AI collaborator at Shekou International School, I will share our approach to supporting a year-long initiative to improve the validity and reliability of summative assessments. We will examine how meeting contributions feed into project records and follow-up, and the corrections needed to preserve agreed intent as the work develops. The case also raises questions about where an agent can act independently and where leadership judgment remains essential.

Participants will work through a short case, distinguish decisions from suggestions, and identify matters requiring a leader’s attention. They will then develop working instructions for an AI collaborator supporting an initiative in their own setting, with a partner introducing a change that tests those instructions.

Participants will leave with a draft collaboration plan and a bounded trial for evaluating the support an agent provides, including the effort required to direct and correct it. No coding experience is required.

LEARNING OBJECTIVES

By the end of the workshop, participants will be able to:

Identify a coordination problem within a school-improvement initiative that could benefit from sustained agent support.
Define responsibilities and decision points for the leader and agent, including when the agent should seek clarification.
Draft and test reusable working instructions, with a practical way to evaluate whether the collaboration helps.

ONE-HOUR WORKSHOP OUTLINE

0–5 minutes: Opening story and participant reflection on an initiative they repeatedly have to reconstruct before moving forward.

5–17 minutes: Goal 1 case, connecting meeting evidence, decisions, project records and follow-up, including a documented correction.

17–32 minutes: Paired exercise using a fictional meeting extract and an agent’s proposed update to identify unsupported commitments, unresolved questions and appropriate next actions.

32–47 minutes: Draft working instructions for one initiative; partners introduce a changed deadline or ambiguous suggestion to test them.

47–55 minutes: Revise the instructions and define a small trial, including signs of useful support and excessive supervision.

55–60 minutes: Discussion and reflection on where recovered attention could make the greatest educational difference.

Prepared examples will allow everyone to participate without configuring software. The take-home planning sheet will capture the initiative’s purpose, essential context, delegated work, matters requiring judgment and trial criteria.