Delegate more without creating another job to manage, Complete Resource Pack

Ideas to keep

1. 01In a prepared example, Victor set up several AI prompts to summarize project threads, draft weekly status reports and flag schedule risks across his department.
2. 02By Friday afternoon he spent more time comparing 3 conflicting drafts and hunting down outdated numbers than he used to spend writing the update himself.
3. 03More agents or more completed actions do not prove a better workflow. Speed of generation means very little if the output leaves you supervising and cleaning up.
4. 04Worthwhile delegation is judged by your total effort and the quality of attention it leaves you with.
5. 05For Victor, useful delegation means one clear summary to decide from, a clear reason for any item returned to him, and final authority over consequential actions.

Try it (optional)

1. Pick one recurring job in your week where you are tempted to add more automation.
2. Estimate the minutes you spend explaining it, reviewing the result and correcting it.
3. Before you widen the setup, ask whether the change would lower that combined effort.

Files and links

Victor is a prepared fictional example written for this course. His Friday afternoon illustrates the problem and is not a recorded AI run.

• Download the Section 6 starter materials (zip)/downloads/leaders/section-06/section-06-starter-materials.zip
• The delegation and governance prepared example/resources/library/leaders/agentic-delegation-and-governance-prepared-example

Ideas to keep

1. 01One chat asked to both build and attack a preferred answer can anchor on it and offer only mild criticism. Two separate chats compared by hand bring back the reconciliation burden.
2. 02A subagent is a separate agent that receives a bounded job from a coordinating agent and returns its findings to it. A skill is a saved set of instructions guiding how one agent does a method.
3. 03The Case-For Subagent gets the shared records plus Maya's preferred phased proposal. The Independent Evidence Subagent gets the same records and criteria without her preferred answer.
4. 04Withholding your preferred answer reduces one source of confirmation bias, though it never guarantees unbiased judgment.
5. 05The coordinator returns 4 sections: where the reviewers agree, where they materially disagree, what evidence is missing, and which uncertainty needs Maya's judgment.
6. 06In ChatGPT Work, subagents are available on eligible accounts, and running several agents at once uses more of your plan than one agent does.

Try it (optional)

1. Pick one high-stakes proposal you want tested before you commit to it.
2. Ask your agent to prepare 2 reviewer briefs for your confirmation, giving your preferred answer to only one of them.
3. Check what each subagent can see, so the independent reviewer cannot read your main conversation.
4. Ask for the 4-section return and read the disagreements first.

Files and links

Maya and Meridian Manufacturing are prepared fictional examples. The reviewer findings and the week-5 conflict are authored teaching material, not output from an observed subagent run.

• Subagent review, in the prepared example/resources/library/leaders/agentic-delegation-and-governance-prepared-example#subagents
• Maya and Alex case files: briefs, rule, routes and effort table/resources/library/leaders/maya-and-alex-delegation-case-files
• Open the interactive delegation and governance page/downloads/leaders/section-06/agentic-delegation-and-governance.html
• Maya's Meridian decision record from Section 3/resources/library/leaders/priorities-and-decisions-prepared-example#record

Ideas to keep

1. 01An action receipt is a saved record showing an action really happened, such as the calendar entry read back after a booking. Check receipts from your own workspace before you grant any rule.
2. 02A standing permission is advance authority for an agent to carry out a narrowly defined routine action within explicit boundaries.
3. 03Alex's rule has 4 parts: the permitted action and destination, the required evidence, routine reporting with readback, and the exceptions that return immediately.
4. 04A broad instruction such as "manage my meeting follow-ups" blurs the line between a private note to yourself and an unreviewed commitment email to a client.
5. 05A clean streak of past runs informs your confidence but never guarantees future behaviour. Your authority always includes an immediate off switch.

Try it (optional)

1. Pick one small routine action you already confirm by hand every week.
2. Review the receipts from earlier runs in your own workspace.
3. Ask the agent to propose a 4-part rule for your confirmation.
4. Note the condition, such as a busy quarter-end, under which you will pause it.

Files and links

The Northstar reminders are a prepared example, so no real receipts sit behind them. Set a rule like this only after checking receipts from your own workspace.

• Standing permissions, in the prepared example/resources/library/leaders/agentic-delegation-and-governance-prepared-example#permissions
• The Monday reminder as Proposed only, Section 1/resources/library/leaders/northstar-follow-through-prepared-example#actions
• The calendar reminder actions skill/resources/library/leaders/calendar-reminder-actions-skill

Ideas to keep

1. 01Two questions sort this out: where does the task run, and what can it reach?
2. 02A local task runs in ChatGPT on your computer and uses the folders you approved there. A cloud task runs on OpenAI's computers and keeps going after you leave. You follow up from the web or your phone.
3. 03With local computer access, a cloud task can reach files on your computer while it is connected, switched on and online. In a company workspace the person who manages ChatGPT usually turns it on.
4. 04A dot is a personal agent with its own cloud computer and browser, on some plans and not in every country. As of 20 October 2026, Pro plans in the European Economic Area, the UK and Switzerland do not include dots.
5. 05A cloud service such as Google Drive linked to ChatGPT is called an app, and this course says linked app. Appearing in ChatGPT's list is not your employer's approval to link company documents.
6. 06A task begun on the laptop cannot switch to the cloud partway through, so Maya starts a new cloud task before she leaves.

Try it (optional)

1. Before you leave, check whether your task runs locally or in the cloud, and whether your computer stays connected.
2. Check that the route can read the specific files and the current decision record it needs.
3. Check your organization's policy on which linked apps are approved for that material.
4. If policy blocks every route, prepare a permitted, non-confidential summary in your approved mobile workspace before you go.

Files and links

Plans, countries and features change often, so check what your own ChatGPT screen shows. Maya's trip is a prepared example, not a recorded session.

• Away-from-desk routes, in the prepared example/resources/library/leaders/agentic-delegation-and-governance-prepared-example#routes
• Maya and Alex case files: briefs, rule, routes and effort table/resources/library/leaders/maya-and-alex-delegation-case-files

Ideas to keep

1. 01Use a recurring job you already noticed while following the course. It might use Meetings and follow-through, Priorities and decisions, or Communication and people leadership.
2. 02When you spot a useful correction, ask the agent to capture it into the skill. When the result already works, ask it to confirm what made the method succeed.
3. 03Counting only the 20 seconds of generation hides the 25 minutes spent hunting files, explaining background, reading and fixing.
4. 04Compare 4 buckets: initial setup, task explanation, output review and manual correction. Keep the one-time setup separate from the recurring weekly effort.
5. 05In Alex's made-up figures, the workflow saves more than half an hour a week and the one-hour setup pays back within 2 weeks, so Alex keeps it.
6. 06One run shows how the workflow performed on this task, not a guaranteed saving every week.

Try it (optional)

1. Choose one recurring job and one of the three course workflows.
2. Run it once on permitted material and capture any correction into the skill.
3. Write your own minutes for the 4 buckets, before and with the workflow.
4. Decide keep, refine or stop, and share only what one proven job needs.

Files and links

Alex's minutes are made up for this prepared example. Replace them with your own figures when you test your workflow.

• Capstone effort table and sharing package, in the prepared example/resources/library/leaders/agentic-delegation-and-governance-prepared-example#capstone
• The meetings and follow-through workflow/resources/library/leaders/meetings-and-follow-through-workflow
• The priorities and decisions workflow/resources/library/leaders/priorities-and-decisions-workflow
• The communication and people leadership workflow/resources/library/leaders/communication-and-people-leadership-workflow
• The meeting follow-up skill/resources/library/leaders/meeting-follow-up-skill

Ideas to keep

1. 01Sections 1 and 2: an agent turned a transcript and your stated purpose into notes and drafts, and a project and a skill stopped you explaining the same job from scratch.
2. 02Sections 3 to 5: an attention queue and a decision record kept choices clear, a voice skill and spoken rehearsal kept messages yours, and prepared observations and open questions kept people conversations fair.
3. 03Section 6: subagents, standing permissions with clear limits, checked away-from-desk routes and a complete-effort comparison let you delegate larger jobs without a second job to manage.
4. 04None of it needed code, because each setup step started with you explaining the work in plain words.
5. 05The principle: you lead the work, and the agent prepares it.
6. 06Completing a course or saving a skill is never a substitute for verifying results in your own workplace and checking receipts.

Try it (optional)

1. Pick one meeting this week whose follow-up you usually end up chasing.
2. Run the meeting follow-through workflow on its transcript.
3. Judge the result with the keep-or-change test from the first section.

Files and links

The 4 questions for any new problem: What result would help the next decision? What grounded context and methods does the agent need? Which exceptions or conflicts must return to me immediately? Which human judgment or conversation belongs only to me?

• The principle and the 4 questions, in the prepared example/resources/library/leaders/agentic-delegation-and-governance-prepared-example#principle
• The meetings and follow-through workflow/resources/library/leaders/meetings-and-follow-through-workflow
• Section 1 resources: Finish the meeting with less work left to carry/resources/library/leaders/finish-the-meeting-with-less-work-left-to-carry-complete-resource-pack
• Section 2 resources: Stop explaining the same work from scratch/resources/library/leaders/stop-explaining-the-same-work-from-scratch-complete-resource-pack
• Section 3 resources: Choose what matters when everything feels urgent/resources/library/leaders/choose-what-matters-when-everything-feels-urgent-complete-resource-pack
• Section 4 resources: Send messages your colleagues recognize as yours/resources/library/leaders/send-messages-your-colleagues-recognize-as-yours-complete-resource-pack
• Section 5 resources: Prepare for conversations you cannot afford to avoid/resources/library/leaders/prepare-for-conversations-you-cannot-afford-to-avoid-complete-resource-pack