Most leaders who finish a management book meet the same problem the next morning: a blank page. The book explains how to write a role charter for an AI agent, and the page in front of them is still empty. So I put the instruments where many executives already work. The E-mployee Operating Model is now a plugin in ChatGPT’s plugin directory, listed under Business & Operations, with me as its developer.
Open the E-mployee Operating Model plugin in ChatGPT →
The plugin also has its own page, with every workflow, example prompts and support: jonahtebaa.com/e-mployee-plugin.
What the plugin does
Each workflow follows one instrument from The E-mployee Operating Model, in the order the book teaches them. It asks only for the facts it needs, marks anything you do not know as missing instead of guessing, and finishes with an artifact a leadership team can take into a decision meeting.
| Workflow | The question it answers | Chapter |
|---|---|---|
| AI access audit | Is our AI rollout producing accountable work, or only logins? | 1–2 |
| E-mployee inclusion test | Does this agent qualify as an e-mployee role: include, defer or exclude? | 3 |
| E-mployer capability audit | Do our leaders have the five capacities needed to direct AI work? | 4 |
| Outcome-first design | What outcome are we buying, before we choose a tool? | 5 |
| Role Charter | Which 13 fields must this role have written down before it starts? | 6 |
| Value Ledger | What does the role really return once review effort, risk and distribution are counted? | 7 |
| First-role selection | Which AI role should we commission first, using the Opportunity Matrix? | 8 |
| Autonomy review | Should this role be promoted, held or demoted on the Autonomy Ladder? | 9 |
| Authority Matrix | Which actions may the role take alone, which need approval, and how is a power revoked? | 10 |
| Incident accountability | Who answers when AI output is wrong, and what changes afterwards? | 11 |
| Risk classification | How much control does this role need, in proportion to its risk? | 12 |
| Human + AI org chart | How does the human organization chart change around AI roles? | 13 |
| Portfolio review | Which AI roles do we keep, rescope or retire? | 14 |
| 90-day roadmap | What are the gates for the first role's first 90 days? | 15 |
A fifteenth workflow, Get started, asks what you are trying to decide and sends you to the right one.
Three ways to start
“We gave everyone AI and nothing changed”
Run the AI access audit. Add the plugin, then ask it, for example: We gave AI tools to 200 staff. Audit whether it is creating real work or just logins. It separates access from accountable work, which is the distinction most rollout reports skip.
“Which AI role do we start with?”
Run first-role selection: Help me choose our first AI role: invoice exceptions, support replies, or market research. The Opportunity Matrix weighs value against how well the work can be bounded and reviewed, so the first role is one you can actually judge.
“This agent is ready for more”
Run the autonomy review, then the Authority Matrix. Giving an AI system more authority should be a decision with evidence, recorded with the conditions under which it is taken back.
What it does not do
It does not connect to your systems. In OpenAI’s terms it is a skills-only plugin: OpenAI describes skills as instructions that teach ChatGPT how to complete repeatable workflows, while live data and actions come from a connected server, and this plugin has none. It is a set of written workflows that run inside ChatGPT, with no server, connector or tracking of its own, and nothing you type reaches me; the details are in the privacy section for the plugin. The instruments are management tools I created. They are not legal advice, a compliance certification or a guarantee, and the workflows say where legal, privacy, security or labor specialists must review.
A real role, run on the same instruments
Most cases in the book are illustrative composites. To show the instruments on a live system, I wrote up the e-mployee I direct every day: Brian, my AI chief of staff, with his full 13-field role charter, his authority limits and four incident reviews.
If you run one workflow this week, run the Authority Matrix on the AI agent you trust most. The gaps it finds are the expensive ones to discover later.
