You have opened the free courses, watched the videos and read the vendor pages, and you still cannot say what you personally should learn next, in which order, and what you can safely skip. The research is done for you here. Each level names what you must understand, what you can ignore at that level, the tools, the risks that enter, and the proof that you have reached it.
Answer for how you actually work this month, not how you would like to. The result points you to the level you are at and the page for the one above it.
When you ask an assistant for help with a piece of work, do you give it the actual source document and a full brief?A full brief says who it is for, what it is for, the constraints, the format and what good looks like. Then you check figures and facts before using the result.
Do you have a Project, custom instructions or a Skill set up, so that your reference material and rules load every time without being re-explained?Claude Projects and Skills, Copilot Notebooks, Gemini Gems, or custom instructions in any assistant.
Have you handed a whole multi-step job to an agent, with stopping rules, and checked its output against cases you already knew the answer to?Research mode, a browser agent, a Copilot or Claude agent, an approved MCP connection.
Levels are per person. In a team engagement each group is placed separately: a fundraising team at level 2 and a leadership team at level 4 are trained separately. For an organisation-wide picture, take the free readiness check.
Days are delivery days from the level below. Percentages are planning assumptions on the tasks in scope, not promises: level 3 sits at the lower edge of the published field studies because verification is added to the work, levels 4 and 5 are extrapolations that each engagement replaces with its own timing, and levels 6 and 7 pay off through other people and through volume. The sources are on the training page.
Has not used an assistant for work. Nothing to unlearn.
Asks for a summary, a search or a first draft, without a method. Results vary by person and by day.
Writes a full brief, grounds it in the real document, verifies before anything leaves the team. The floor.
Loads context once (Projects, custom instructions, Skills), so nothing is briefed twice and quality stops depending on who asked.
Hands whole multi-step jobs to agents, supervises rather than types, runs workstreams in parallel.
Builds Skills, agents and connections for colleagues; reviews their output; owns the operating guide.
Runs agentic work on folders of files from the terminal (Claude Code), without writing code.
Why the days are not linear: level 3 is the longest step because it carries all the background (what a model is, the brief, the checks, the data rules). Every later level assumes it and adds one capability.
Then 2 more to level 4, 1.5 to level 5, 1.5 to level 6, 2 to level 7. Ten delivery days in all from level 2 to level 7.
Groups of up to nine, trained on your own tasks and documents. Online or in person; in person, we arrange the venue, hotel and hospitality.
You receive a written quote after a free scoping call, whether you come as an individual, a representative or a team. Book the scoping call →
Your own playbook: the tasks to hand over, the ones never to, the prompt patterns that worked on your files, the four checks, and what to log. Plus an attendance record for your evidence file. See a sample playbook →
The method follows the four competencies of the AI Fluency framework published by Anthropic with Rick Dakan and Joseph Feller source. The framework is public, so the course can be checked against it.
Understand, do, tools, the risks that enter at that level, what to skip for now, the proof, a printable cheat sheet, example prompts for the work you actually do, the day-by-day plan and the days it takes.
What a model is, the context window, model tiers, your data rules. The full brief, the source document, the four checks. Prompts for a donor report, a policy brief, a literature matrix.
Read level 3 → Level 4Context engineering. Projects, custom instructions, Skills, memory, one read-only connector. Build a set-up a colleague can use without you.
Read level 4 → Level 5Agentic AI, deep research, browser agents, MCP at the user level, blast radius. Brief a whole job with stopping rules and a test set.
Read level 5 → Level 6Design Skills and agents for someone else, with an owner and an operating guide. Evaluation, injection and exfiltration, choosing vendors.
Read level 6 → Level 7The terminal, Claude Code without code, CLAUDE.md, hooks, git, scheduled runs. Batch work on hundreds of files, and what is still engineering.
Read level 7 → IllustratedIllustrative personas from NGOs, agencies, a foundation, a university and a social enterprise climbing the ladder, level by level.
See the journey →Thirty minutes is enough to place you or your team on the ladder and say honestly whether a course is the right next step.