Free course
From Prompt to System
Pick one job in your business and finish with AI doing it.
- 56 lessons
- Four missions
- 11 hours of lessons
- About 30 hours with the builds
- A certificate
Free account. Email or Google, about a minute.
The syllabus
56 lessons in six stops.
Every lesson ends with a check. The job you pick in Launch is the one you build in every mission after it.
Launch The mission
What a model actually does with your words, and the one job in your business you will carry through every mission.
Part 1
The mission
What a model does with your words, and the job you will build in every mission after this one.
Mission 1 Make it answer
Five parts, in a fixed order, that turn a vague request into an instruction a model can follow. Then the part almost nobody does: twenty real inputs you labelled by hand, and a prompt you can change without guessing whether you made it worse.
Part 1
Why it failed
What a model actually does with your words, and the four ways a request goes wrong before the model ever sees a problem.
Part 2
The five parts
Context, instructions, format, rules, examples. This is the one that takes a second read, and every lesson after it is these five again.
Part 3
Judgement, not prose
Prompting for a decision rather than for writing, which is where most of the money is and where most prompts fall apart.
Part 4
Proving it
Twenty cases you labelled by hand, reading your misses out loud, and a prompt you can change on purpose.
Part 5
Judgement
Three prompts scored against the seven lines, so the score means the same thing when you run it over your own.
Mission 2 Make it run
From nothing to two workflows running on your own accounts. Triggers, JSON without fear, branching, the node that lies to you, and the real monthly bill for self-hosting against cloud.
Part 1
The sentence
What an automation actually is, and the three clauses every one has.
Part 2
The moving parts
Triggers, data and the shape it arrives in. This is the one that takes a second read, and it is the one that stops you guessing later.
Part 3
The first build
A form that files itself, running on your own accounts by the end of the lesson.
Part 4
When it does not work
Reading an error, the node that reports success while doing nothing, and calling an API you have never seen before.
Part 5
The second build, and the bill
A digest that writes itself every Monday, what a month of it costs, and where to put it so it keeps running.
Part 6
Judgement
Three workflows scored out loud against the seven lines, so the score means the same thing when you run it over your own.
Mission 3 Make it act
Three agents that do real jobs: one triages your inbox, one researches your leads, one writes your weekly report. Then seven lines you can score your own directive against, and the shape to build the next agent without help.
Part 1
The model
What an agent actually is, and why the job you picked is worth giving one.
Part 2
The instructions
Writing directives an agent can follow, and the architecture they sit in. This is the one that takes a second read.
Part 3
The first build
An agent running on your own inbox by the end of the lesson, and the hands it needs to act on anything else.
Part 4
Two more, and proof they work
A research agent, a report generator, and the testing that catches a break before a client does.
Part 5
Keeping it alive
What breaks in production, and what a run costs.
Part 6
Judgement
Score a set of instructions against a published checklist, pick where the fourth agent should run, and find a fault by hand when it breaks.
Mission 4 Make it pay
Ten systems somebody is already paying a person to do badly. For each one: the pain in the buyer's words, the deliverable they can see without logging in, the build, and what it is worth worked out rather than guessed. The last lesson prices an eleventh the same way, and then yours.
Part 1
The idea
Why nobody buys an automation, and the arithmetic that turns a build into a number. This is the one that takes a second read, and every lesson after it uses it.
Part 2
Field guide: the first five
Triage, research, proposals, collections and repurposing. Each one priced from the method rather than from a guess. Read the ones closest to your job.
Part 3
Field guide: the second five
Reputation, onboarding, reporting, competitor watch and recruiting screens.
Part 4
Judgement
Three pitches scored against the seven lines, and choosing which of the ten fits the situation you are actually in.
Part 5
The eleventh is yours
Pricing the job you have built since the second lesson, with the same method as the ten.
Mission review Your mission, scored
Your job's system scored against all four seven-line rubrics, the certificate, and what to build next.
Part 1
Your mission, scored
All four rubrics over your own system, the certificate, and the next job.
What you build
Nine builds, and you keep all nine.
Each one is built in the lesson named on it, on the job you picked, and it is yours when the course ends.
Build 1
Your job, written down
The one job in your business you build in every mission, with how often it happens and what it costs today.
Build 2
A prompt that works first pass
One real prompt for a job you actually do, written to the five parts and scored against the seven lines.
Build 3
Your twenty case set
Twenty real inputs with answers you agree on, which is what lets you change the prompt without guessing.
Build 4
The form that files itself
A submission becomes a row, a renamed file and a message, with nobody copying anything.
Build 5
The Monday digest
Three sources, one summary, in your inbox before you open your laptop on Monday.
Build 6
Email triage agent
Routes, drafts, archives, and pings Slack. Runs on your own inbox from Mission 3, lesson 4.
Build 7
Lead research agent
A company domain goes in. A scored one-page brief with 3 talking points comes out.
Build 8
Weekly report generator
Pulls the week's numbers from your sheet. Writes the Monday report before you wake up.
Build 9
Your job, priced as a system
The job you built across four missions, priced from its own arithmetic and pitched in the buyer's words.
Questions
Before you start.
- Do I need to code?
- No. Mission 1 runs in a chat window on a free tier, Mission 2 on a no-code platform, and the code path in Mission 3 is optional.
- How long does it take?
- The lesson time is at the top of this page. Budget about thirty hours with the builds, and most of that is building your own job.
- Which model should I use?
- Whichever you already have. Mission 1, lesson 1 shows you how to tell a prompt problem from a model problem, which is the only time the answer matters.
- What will this cost me?
- The free tier of either platform carries both Mission 2 builds, and a few dollars of model calls covers every build in Mission 3. Mission 2, lesson 11 works out what a real month costs once you run more, from your own numbers rather than ours.
- Is it really free?
- Yes. Every lesson, quiz, and certificate. It asks you to make a free account, which is what saves your progress and puts your name on the certificate. Advizr makes money building these systems for clients, and the academy is how we prove we know how.
- Can I skip a mission?
- You can open any lesson once you enroll. Each mission builds the job from the one before, so the order is the advice. The certificate needs every lesson outside the field guide and every check passed.
- What is the certificate worth?
- A public page with your name and the course, earned by finishing every lesson outside the field guide and passing its check. It is not an accredited credential.
- Do I need an account?
- Yes, and it is free. You sign in with your email and a code we send to it, so there is no password to remember.
The certificate
Yours when every check is passed.
Finish the 46 lessons outside the field guide and pass their 46 checks. The field guide is yours to read and counts toward nothing. The certificate carries your name, every lesson and every score, and the course’s four missions as its seal.
How anyone checks itAwarded to
Your name here
for finishing From Prompt to System
- 46 of 46 lessons
- 46 of 46 checks passed
- 100% of answers right
- 9 builds
A specimen, issued to nobody. Yours carries your name and your scores.