The training program behind every KeyPlayers assistant
School of AI
Opens October 7The model landscape, prompting like a senior EA, Claude and ChatGPT in depth, briefing and supervising agents safely, and AI playbooks for your role. Ends in a graded week as a founder's AI-powered EA.
This program opens on Wednesday, October 7. Applications open the same day.
Assistants who hold the KeyPlayers AI-Trained VA and want to do this work for clients.
By application, after KeyPlayers Foundations. The admissions team reviews every application.
Every graduate has already done this work once, under grading. The final asks them to produce:
Curriculum
Modules pair lessons with quizzes. Modules marked Drill include hands-on practice on a realistic task. Downloads are available to trainees once they sign in.
How AI models actually work, what Claude, ChatGPT, Gemini, Copilot, Grok, Perplexity, Viktor and the automation agents each do best, and how to pick the right tool for every task inside the client's approved stack and data settings. Current as of September 2026.
Assessment
A realistic case, worked the way a client would expect. A KeyPlayers trainer grades it against the rubric below. AI may prepare first-draft scores backed by exact quotes from the work, and the trainer decides every result.
Run a founder's week as the AI-trained EA for Tessa Marchetti, founder of Copperleaf Studio, a 12-person design studio in Calgary: build her context pack, choose the right tool (or no AI) for each task, write reusable prompts, review and rebrief an AI agent's onboarding plan, turn a Claude Research report into a checked call brief, write the Friday update, map where AI can take work off the founder's plate, and log every AI use. Along the way: a request to paste client data into a personal free AI tool, an agent plan that would email a client without approval, a hidden instruction in a client's form, an invented statistic, a Slack phishing message and a vague instruction that needs a clarifying question.
A program opens to trainees only once each review below is signed off. Every sign-off records exactly which version of the content was reviewed, and lasts 12 months, so the program is reviewed again every year.
Status: Opens October 7
By application, after KeyPlayers Foundations. The admissions team reviews every application. New to KeyPlayers? Take the skills quiz, then apply. Already training with us? Open the program from your Academy account.
The model landscape, prompting like a senior EA, Claude and ChatGPT in depth, briefing and supervising agents, and AI playbooks for every seat.
Browse every schoolFor assistants
Nine real workplace situations, with your score and what each answer means. Training is free for KeyPlayers candidates.
Files open for trainees inside the Academy after sign-in.
The advanced prompting course for KeyPlayers EAs. The seven-part KeyPlayers prompt method, a reusable context pack for every client, prompts for real EA work (triage, voice drafts, prep, research, summaries, SOPs, voice notes), iterating with follow-ups, chains, files, spreadsheets and agents, and red-teaming every output before it reaches the client.
Files open for trainees inside the Academy after sign-in.
Side-by-side mastery of the two assistants your clients pay for. Set up a client workspace in either tool, connect it to Gmail, Calendar and Drive with the least access that works, make documents, spreadsheets and pages, run research you can defend, and turn recurring work into a skill or a custom GPT. Written from the assistant's seat: what AI does, what stays with you, and what needs the client's yes.
Files open for trainees inside the Academy after sign-in.
The advanced agent course for executive assistants. What an agent is and how far to let it act, which agent to use where (Claude, ChatGPT Work, Copilot Cowork, Gemini, Viktor, Grok Bot, Zapier, Make and KeyCommand), how to brief one, how to supervise its plans, approvals and logs, how to move a manual workflow onto an agent safely, and how to handle prompt injection, over-permissioned connectors, spending, sending and data leakage when something goes wrong.
Files open for trainees inside the Academy after sign-in.
Leverage from where you sit. Audit a week of your own work, then run role-by-role playbooks for the founder's EA, admin and support, sales and marketing, and specialist seats in bookkeeping, mortgage and real estate. You finish with a personal AI operating system and a 30-day plan your client can approve and you can measure.
Files open for trainees inside the Academy after sign-in.
Every request gets the right call for the EA and for any agent: act when it's routine and theirs, ask first when it commits Tessa or leaves Copperleaf, and stop and escalate injections, data slips and anything unexpected. Regulated questions go to the licensed professional.
Data classes are applied before every AI step: nothing in a personal tool, nothing Pinecrest in any AI tool, Restricted details removed, prospect data kept private. The phishing message is reported with specific red flags and not acted on.
The context pack gives any approved AI tool what it needs to work for Tessa, uses current rules, and keeps personal and Restricted details out. Prompts have a role, task and context, format and constraints, use placeholders, and treat content as data.
Each task gets the right tool, or no AI, for its shape and data class: fast models for high-volume summaries, the most capable or default model for research and synthesis, the agent for multi-step work in connected tools, and by hand where AI isn't allowed. Only approved tools.
The agent's plan is reviewed step by step before anything runs. The hidden instruction is recognized as prompt injection and escalated the same day. The new brief says what the agent may and may not do, stops before anything external and asks for a list of changes.
Every claim in the AI research report is checked against its source before anything reaches Tessa. Invented figures and phantom sources are removed, wrong facts corrected, unneeded personal details dropped, and Tessa is told why.
Vague instructions get one clear question with options before anything is done. Messages to Tessa are short, labelled and easy to answer. Drafts in her name match her voice and rules. Hard conversations (the statistic, Hollis) are kind, factual and firm.
Recurring work is mapped to what AI does, what the EA does and what the founder approves, with the tool and data class, and a way to measure the result with real timings. Tool questions get a reasoned recommendation. The AI log is honest, complete and shows what stayed out of AI.
Written feedback on every result. A result can be appealed within 7 days, and a different trainer decides the appeal.
For businesses
KeyPlayers assistants train and are graded here. Ask us which programs fit your business and how we match you with the right person.
