Take one recurring task from your actual job through the full white-belt cycle: select it deliberately, frame it properly, verify it honestly, and save the result as your first reusable asset. This is real work shipped, not an exercise.

Pick a recurring weekly task that is tedious and instantly verifiable (module 1's strengths map is your guide)
Write the one-sentence spec: 'I need [output] for [audience] that achieves [outcome]'
Build the full prompt with all five building blocks (goal, role, context, format, constraints) and label them in your draft
Run it, then apply three rounds of targeted feedback naming what to keep and what to change
Assign the stakes tier and run the matching verification: check every fact, name, and number against what you know or hold
Use the final output for real, send it, file it, present it
Convert the winning prompt into a bracketed template with a usage line and a verification line
Log the honest time spent versus the manual version, including verification

DeliverableA piece of real work produced with AI, verified and actually used, plus your first saved template with its verification note, and a time log proving (or disproving) the payoff.

Brown Belt

Advanced

Not started

Pack the Context

Close the context gap. Feed the AI what it cannot possibly know (your background, your examples, your voice) and turn generic output into work that sounds like you.

~30 min (est)4 prompts 4 intermediate

Decide What to Delegate

Audit your working week, sort tasks into automate, augment, or keep, and pick first delegations that build capability instead of risk.

~30 min (est)4 prompts 4 intermediate

Verify What You Cannot Personally Check

Master the hard case: techniques for checking AI output in domains where you are not the expert, from triangulation and source demands to consistency probes and adversarial passes.

~30 min (est)4 prompts 4 intermediate

Handle Company Data Safely

Know what never goes into a prompt, master the anonymisation techniques that let you get help anyway, and audit the data controls of the tools you actually use.

~30 min (est)11 prompts 4 intermediate · 7 advanced

Iterate Like an Editor

Treat first drafts as raw material: give targeted feedback instead of regenerating, request true variations, and learn to push back against an assistant that agrees too easily.

~30 min (est)4 prompts 4 intermediate

Put AI to Work on Documents

The knowledge worker's bread and butter: purposeful summaries, structured extraction, version comparison, and the chunking technique that beats long-document limits.

~30 min (est)4 prompts 4 intermediate

Build Your Prompt Playbook

Turn your wins into infrastructure: templatise the prompts that work, attach a verification step to each, and package the playbook so your team benefits too.

~30 min (est)4 prompts 4 intermediate

Use AI With Numbers, Carefully

Language models predict text, they do not compute. Learn to get the method instead of the answer, verify data interpretations against the table, and run a sanity-check ritual on every figure that travels.

~30 min (est)4 prompts 4 intermediate

Capstone · Brown BeltKnot open

Your Personal AI Playbook

Build the operating asset that compounds: a tested, safety-checked playbook of five templates covering your most frequent AI-suitable work, with context blocks, verification steps, and one template shipped to a colleague.

Run the delegation audit on your full working week and pick the five highest-value AUTOMATE and AUGMENT tasks
Build a bracketed template for each, covering at least three task types (writing, summarising, extraction, numbers, thinking support)
Write your standing context block and a project block, and note which templates need them pasted first
Run the data-safety pass: classify what each template touches and anonymise or re-route anything above internal tier
Attach a one-line verification step to every template, matched to its stakes tier
Stress-test each template against a messy input and an edge case; patch and version what breaks
Assemble everything on one page titled 'AI Playbook v1' and store it where your work lives
Ship your best template to one colleague with a worked example, and fold their confusion back into the template

DeliverableA one-page, five-template playbook (tested, data-safe, verification-noted) in daily use, with at least one template adopted by a colleague.

Black Belt

Mastery

Not started

Design an AI-Assisted Workflow

Move from single tasks to systems: decompose a recurring process, assign each step to AI or human, place verification checkpoints where errors are expensive, and document it so it survives you.

~35 min (est)5 prompts 5 advanced

Keep Your Judgment Sharp

Counter automation complacency before it costs you: blind tests, error logs, think-first discipline, and using AI as a sparring partner that strengthens your reasoning instead of replacing it.

~35 min (est)4 prompts 4 advanced

Diagnose and Fix Bad Output

Stop random retries: classify failures into the five modes, apply the matching fix, know when to restart a degraded conversation, and when to stop and hand the task to a person.

~35 min (est)24 prompts 24 advanced

Work Responsibly and Stay Current

Own what ships, know when to disclose, audit for bias, and build a lightweight routine that keeps your skills current as models change. Then pick your graduation path into the tool dojos.

~35 min (est)9 prompts 9 advanced

Capstone · Black BeltKnot open

The Workflow Redesign

Redesign one recurring process end to end as an AI-assisted workflow with human checkpoints, run it live twice, measure the gain honestly, and document it so a colleague could run it without you. Portfolio-worthy operator work.

Choose a recurring process that consumes at least two hours a week and map it into five to eight steps with inputs and outputs
Assign each step AUTOMATE, AUGMENT, or KEEP (decision steps stay human) and override at least one AI suggestion with reasoned judgment
Design two verification checkpoints with factual pass rules at the highest-consequence points
Attach a tested template from your playbook to every AI step, including any numbers step routed method-first through a spreadsheet
Run the workflow live on two consecutive real instances, enforcing both checkpoints and logging time, errors, and friction
Diagnose the weakest step with the five failure modes and redesign it; version the workflow
Run the responsibility pass: data classification per step, disclosure decision, and a bias check on any people-affecting output
Write the one-page runbook (steps, templates, checkpoint rules, timings) so a competent colleague could run it tomorrow
Present the before-and-after numbers to your team or manager, with the runbook attached

DeliverableA documented, twice-run AI-assisted workflow with measured time savings, enforced checkpoints, and a runbook a colleague could execute, plus the before-and-after case you presented.

Beginner Intermediate Advanced Bar length is the module's prompt count against the longest module here; its segments are that module's own mix. Minutes are the curriculum's per-module estimates, summed over what you have woven.

One cell is one module, at its belt cloth Woven cell: module complete Knot open until the belt's last module is woven. Kept on this device only.

// SOURCES

Curriculum Record

Every published source this curriculum cites: 18 distinct references, counted from the curriculum brief and from the prompts themselves. Prompts that quote one carry a source line; the rest are marked unsourced author prompts.

01dol.govAttachment I (Accessible PDF)
02teachingcommons.stanford.eduunderstanding-ai-literacy
03arxiv.org2412.12107v1
04tandfonline.com10494820.2025.2514372
05pwc.comai-hallucinations
06ibm.comai-hallucinations
07natlawreview.comai-hallucinations-are-creating-real-world-risks-businesses
08hai.stanford.edube-careful-what-you-tell-your-ai-chatbot
09sosafe-awareness.comai-prompting-mistakes-that-risk-company-data
10resources.rework.comai-literacy-the-new-workplace-skill
11it.nc.govdos-and-donts-using-artificial-intelligence-responsibly-and-safely
12academy.techpresso.coai-best-practices
13nhbr.comworkplace-ai-the-simple-rules-everyone-must-learn
14support.microsoft.comsafety-tips-for-using-ai-at-work
15microsoft.comai-data-security-in-workplace-best-practices
1624hourtek.comuse-ai-tools-safely-at-work
17firmspace.combefore-you-hit-enter-5-ai-usage-tips-to-protect-your-data
18grow.googleresponsible-ai-best-practices

AI Foundations Dojo

Untied

Fifteen modules

Curriculum Record

Get the morning Signal