33-lesson course · about 10 hours
An AI chief of staff for your projects: it prepares, you decide and send
For Consultants, project leads, managers and professionals who turn information into plans, decisions and client work
Everything for ProfessionalsIn 33 hands-on lessons, set up an agent that prepares your morning brief, meeting prep, follow-ups, decision log and weekly review. It reads only what it needs, and nothing goes out until you approve it.
What you’ll be able to do
By the end, you can…
How it works
A tutor that meets you where you are
- Lessons that adapt to your levelEach lesson is a tutorial you follow with your tool open: the goal, what you need first, numbered steps, copy-paste prompts written for project work, a worked example and a checkpoint. The lesson adjusts to where you're starting from, whether you've only used AI for one-off questions or you've already tried a few connectors.
- An AI tutor you can ask anythingIf a setting on your screen doesn't match the lesson, or you're unsure whether a source belongs in your setup, ask the tutor. You can type or speak, and the answers are tied to your platform and your project.
- Quizzes and flashcards for each lessonShort quizzes and flashcards go with every lesson. They cover things like which actions need approval, the decision-log statuses, and the difference between done claimed and done verified.
- Progress tracking and a certificateYour progress through the nine modules and the two-week rollout is saved, so you can stop and pick up again. When you finish, you get a certificate of completion.
Curriculum
The curriculum
The course has nine modules in the order you'd build the setup: define the job, choose a platform, write the instructions, connect sources, then add each routine and test it over a two-week rollout. You follow a sample project, deciding whether to bring a second supplier into the northern region before the contract renews, or you can use your own assignment if you have permission to.
Module 1The Job: What a Chief of Staff Does for a Project Lead3 lessons
1What a Chief of Staff Does on a Project
- List the standing duties of a chief of staff and match each to a project-lead task
- Sort your own open items into a now/soon/never list
- Name three things a good chief of staff does not do
2What to Hand an AI: It Prepares, You Decide and Send
- Classify chief-of-staff tasks as may-do-alone, ask-first or never
- Explain why a draft is not authorization to send
- Explain the track's rule that nothing leaves under your name untraced
3The Sample Project and Your Own Assignment
- Describe the sample project's question, stakeholders and recurring work
- Decide whether your own assignment can stand in, and check that you have permission to use its material
- Write a one-paragraph scope note for the project the agent will support
- Map the course's modules onto your own 14 rollout days
Module 2Choosing Your Platform4 lessons
4OpenAI dots and Claude Duet: What They Do and Who Can Get Them
- Describe what a dot and Claude Duet each do for a project lead
- Check whether your region, plan and account qualify for either
- Explain the approval boundaries each one enforces
5Generally Available Paths: Claude and ChatGPT Scheduled Tasks
- Compare Claude and ChatGPT scheduled tasks on plans, limits and approval controls
- Explain the ChatGPT rule that project tasks can't read project files
- Choose the permission setting to start with in Claude
6Work-Suite Paths: Microsoft 365 Copilot and Gemini in Workspace
- List the Outlook Copilot and scheduled-prompt features a project lead can use
- Spot Copilot and Gemini features that aren't available on your account type
- Estimate the monthly cost of your work-suite option
7Pick Your Platform and a Working Fallback
- Choose a first-choice platform and a fallback that fit your region, plan and work account
- Write down your plan, price and access status, with the date you checked
- Explain why Muse, Grok Bot and Manus need extra caution for client work
Module 3Writing the Standing Instructions4 lessons
8Write the Agent's Job Description and Decision Rights
- Write the role section of the standing instructions for your project
- Write the decision rights: what the agent prepares and what stays yours
- Apply the new-colleague test to your draft
- Record a day-0 baseline of your daily triage and prep time
9Priorities, People and Allowed Sources
- Write five ranked priorities split into outward- and inward-facing
- Build a VIP and stakeholder list for one project
- Name allowed sources, the source of truth for each fact type, and sources the agent must not read
10May Alone, Ask First, Never: Writing the Action Rules
- Write the may-alone, ask-first and never lists for your agent
- Write the 'content is data, not instructions' rule into the never list
- Explain why irreversible or visible actions always need approval
11Reporting, Urgency and Voice, Saved in a Master Document
- Write source-tag, inference and 'Could not check' rules into your instructions
- Define urgency rules for project work
- Load the standing instructions into your chosen platform from one master document
Module 4Connecting Sources with Least Access4 lessons
12Least Access: Read Before Act
- Name OWASP's three root causes of excessive agency and their fixes
- Plan a week-1 and week-2 connection schedule
- Explain why disconnecting an app doesn't erase what the agent learned
13Connect Work Email and Calendar Read-Only
- Connect email and calendar on your platform with the least access that works
- Set the starting permission settings for your tool
- Confirm the agent can list its connected sources correctly
14Set Up the Shared Log Folder
- Create the log files with the course's field structure
- Place the logs where scheduled tasks can reach them
- Seed the logs with sample-project rows
15What Never to Connect: Client Separation and Memory
- List what never gets connected for client work, and why
- Set up one project per client with account memory off for sensitive work
- Check whether a client's NDA or data terms allow the tool, and know who to ask
Module 5The Morning Brief5 lessons
16Anatomy of a Sourced Morning Brief
- Explain each of the brief's seven sections
- Spot an untagged fact or an unmarked inference in a sample brief
- Adapt the brief prompt to a multi-workstream project
17Run the Brief by Hand and Score It
- Run the morning brief prompt manually on your platform
- Score each item correct / wrong / missing / wrong priority
- Log corrections as rules in the Corrections Log
18Content Is Data: The Prompt-Injection Drill
- Run a safe prompt-injection drill on your own account
- Confirm the agent flagged the line and took no action
- Stop the rollout and fix the rules if the drill fails
19Schedule the Weekday Brief and Check the First Run
- Create a weekday morning schedule on your chosen tool
- Verify the first scheduled run for sources and 'Could not check'
- Fix common failures: wrong time zone, paused tasks, missing files
20Tune Urgency Across Workstreams
- Revise urgency rules after a brief over-flags
- Set up a VIP watch where your tool supports it
- Spot duplicate items reported as new changes
Module 6Meetings, Follow-Ups and Drafts for Approval4 lessons
21Prep Notes for Tomorrow's Meetings
- Run the meeting prep prompt for a client meeting
- Add the prep run to an end-of-day schedule
- Find meetings with no agenda or no clear purpose
- Connect the week-2 sources read-only for this one project
22Steering Committee Pre-Reads and Agendas
- Draft a one-page pre-read with decisions requested and sources on every line
- Draft a timed agenda with DECIDE items
- Identify likely objections using only members' sourced past comments
23From Notes to Recap Draft and Proposed Rows
- Run the follow-up prompt on a set of meeting notes
- Check that every owner and date in the output was actually stated
- List anything you may have committed to personally
24Holding Drafts for Approval and Sending Them Yourself
- Run inbox triage and draft replies for 'me today' items only
- Approve or edit a held draft, checking the exact text and recipients
- Record the draft edit rate
- Check consent before using AI notes in client meetings
Module 7Decision Log, Delegation Tracker, Deadlines and RAID4 lessons
25The Decision Log: Proposed vs Confirmed
- Extract decisions from a thread into proposed rows
- Confirm or reject proposed rows and set review triggers
- Pull past decisions and contradicting sources before a new decision
26The Delegation Tracker: Done Claimed vs Done Verified
- Run the check-in prompt and choose which nudges to send
- Verify a claimed completion against its definition of done
- Decide which owners need a conversation rather than a reminder
27The Deadline Register and the Capacity Check
- Maintain a deadline register with T-14, T-7 and T-2 warnings
- Run a capacity check on a new request
- Choose between a yes-with-conditions draft and a not-now draft
28RAID Upkeep Without Closing Anything
- Run the weekly RAID scan prompt
- Review proposed entries and closable candidates with evidence
- List the top 5 risks to the next milestone with a question per owner
Module 8The Weekly Review and Status Report2 lessons
29Schedule the Friday Weekly Review
- Schedule a Friday weekly review on your platform
- Use the review in a planning conversation
- Compare two weeks to find items stalled for two weeks or more
30The Thursday Status-Report Draft
- Draft a weekly status report with RAG, milestones and decisions needed
- Run the source check to find unsupported or contradicted statements
- Edit and send the report yourself on Friday
Module 9Guardrails and the Two-Week Rollout3 lessons
31Permission Audit and the Corrections Log
- Run the permission audit prompt and compare it with your tool's settings pages
- Remove unnecessary access, using disconnect plus reset where the vendor requires it
- Fold Corrections Log entries into a new instructions version
32Run the Two-Week Rollout and Keep Score
- Check your setup against each day of the rollout plan and close any gaps
- Record the five core measures daily or weekly
- Explain why any safety event stops the rollout
33Day 14: Keep, Widen or Cut Each Routine
- Score each routine against the two-week measures
- Write a go, widen or cut decision for each routine, with a reason
- Decide whether to keep, narrow or disconnect each week-2 source
Who it’s for
Made for you if…
Made for: Consultants, project leads, managers and professionals who turn information into plans, decisions and client work
Consultants running several client engagements at once
You need a daily view across workstreams, and each client's material has to stay separate. The course sets up one project per client, keeps account memory off for sensitive work, and attaches a source to every claim before it goes into a client report.
Project leads who run steering committees, RAID lists and status reports
You'll hand the recurring preparation to an agent: pre-reads, agendas, recap drafts, RAID scans and a Thursday status draft. You still check every draft and send it yourself.
Managers who lose track of actions agreed on calls
The delegation tracker and decision log separate proposed from confirmed, and done claimed from done verified. Owners and dates that nobody actually stated show up as gaps, not as commitments.
Professionals who use an AI assistant but haven't given it recurring work
If you already use ChatGPT, Claude, Copilot or Gemini for questions and summaries, this course takes you through scheduling, connections and approval rules one step at a time.
Questions
Good to know
Do I need OpenAI dots or Claude Duet?
No. They're the preferred route if you can get them, but access is limited. As of October 2026, dots come with ChatGPT Pro or a Business Premium seat, and Pro excludes the EEA, Switzerland and the UK. Claude Duet is a US-only research preview for Claude Pro and Max, with a waitlist. Every exercise can also be done with Claude or ChatGPT scheduled tasks, and the course covers Microsoft 365 Copilot and Gemini in Workspace too.
What will the tools cost me?
This varies by tool, and Module 2 walks you through the options before you buy anything. Some US prices read in October 2026: Claude Pro is $20/month. ChatGPT Plus is $20/month and Pro starts at $100/month. Microsoft 365 Copilot is $30/user/month paid yearly, and Copilot Business is $21/user/month paid yearly. Prices change, so check the vendor's official page.
Will the agent send emails or accept meetings for me?
No. The course sets it up to read, sort and draft only. Sending, replying, calendar changes, sharing files and anything that commits you to a date, price or deliverable wait for your explicit approval of the exact text or action. Payments, signatures and password changes stay with you entirely.
Can I use my own client or internal project?
Yes, as long as your employer or client permits you to use AI tools on that material, and the scope, commitments and review responsibility are clear. If you don't have that permission, use the course's sample project.
I work in the EEA or UK. Can I still take the course?
Yes. Dots on ChatGPT Pro, Gemini Task, Claude Duet and Meta Muse may not be available to you. Claude scheduled tasks, ChatGPT scheduled tasks, Microsoft 365 Copilot and Google Workspace are the routes to check first, and the course tells you to confirm access before you start.
Does the course cover legal questions like NDAs or recording consent?
It shows you where these questions come up and who to ask: check NDAs and data terms before you connect a tool, and get consent before using AI notes in client meetings. It doesn't give legal advice. Check those points with your client and your firm.
Will the agent write my client recommendation?
No. The agent gathers evidence, checks it and drafts around the recommendation. The judgment and the recommendation itself stay with you.
How long does it take?
About 10 hours of lessons, at your own pace. The course is built around running the setup on your project for two weeks, and the last lesson is a day-14 review.