Foundations pathway · course 2 6 domains 2 × 50-item mocks Academy-aligned

This track is independent preparation for the OpenAI Academy course Applied AI Foundations. It is not an official OpenAI course, not an official assessment, and not endorsed by OpenAI. The material is built from publicly available OpenAI learning objectives and product documentation. For how the credentials fit together, read the credential landscape; for how the assessments work, read the assessment model.

What this track prepares you for

Applied AI Foundations is the second course of the Foundations pathway, after AI Foundations and before Agents and Workflows. OpenAI’s published time for the Academy course is 75–90 minutes, and it is delivered inside ChatGPT. Its three published objectives are to break work into steps, define inputs and outputs, and build a repeatable workflow.

The audience is the giveaway: this course is for knowledge workers who already prompt competently and now need repeatability. You are past “how do I write a good prompt” and into “how do I turn this thing I do every week into something a colleague can run, that produces the same quality every time, that I can improve on evidence.” Passing the Academy assessment (and any mock here) is about judgment on decomposition, contracts, capability choice and oversight, not about prompt wording.

Blueprint

Weights are our design choices for this independent mock, chosen to reflect where the Academy course spends its time. Each mock exam has 50 items.

#	Domain	Weight	Items (approx.)	Track page
1	Finding and Scoping Opportunities	16%	~8	Domain 1
2	Decomposing Work into Steps	18%	~9	Domain 2
3	Inputs, Outputs and Contracts	16%	~8	Domain 3
4	Choosing the Right Capability	20%	~10	Domain 4
5	Review Points and Human Oversight	16%	~8	Domain 5
6	Repeatability and Improvement	14%	~7	Domain 6

Where the marks are

Capability Choice (20%) and Decomposition (18%) are 38% of the mock together. If you learn one thing from this track, make it the capability-selection framework in Domain 4 — prompt vs saved instruction vs Project vs custom GPT vs workspace agent vs API application — and the decomposition patterns in Domain 2. The rest of the track hangs off those two decisions.

Two independent mock exams

This track ships two full-length, domain-weighted independent mock exams. Mock Exam 1 is your diagnostic — sit it first, untimed, to find your two weakest domains. Mock Exam 2 is deliberately harder — more multi-constraint stems and FIRST / BEST / MOST cost-effective / TWO qualifiers — and it is your readiness gate: reach 80%+ timed before you take the real Academy assessment. All items across both mocks and the domain pages are distinct.

The mindset this track rewards

Foundations taught you to treat one output as a draft to verify. Applied AI teaches you to treat a recurring task as a system to design. Correct answers on this track consistently:

Scope before automating — screen a task for frequency, time cost, error tolerance and data sensitivity before building anything.
Decompose rather than mega-prompt — small named steps you can inspect beat one giant instruction that fails opaquely.
Treat every step as a contract — named inputs, required fields, a defined output shape, acceptance criteria, and a rule for missing input.
Reach for the lightest capability that meets the need — a saved prompt before a Project, a Project before a custom GPT, a custom GPT before an API application.
Place review where a mistake would be expensive, not everywhere, and sample rather than gate high-volume work.
Improve on evidence — measured quality and cycle time — not on how the workflow feels.

Wrong answers reach for the heaviest tool, hide everything in one prompt, add a human gate on every low-stakes item (or none on a high-stakes one), and call a workflow “done” when it has never been written down or measured.

Suggested time allocation

A focused 4–6 hour plan including both mocks. Weight your own revision by domain weight × your error rate, not by weight alone.

Domain	Weight	Suggested time
Choosing the Right Capability	20%	75 min
Decomposing Work into Steps	18%	65 min
Finding and Scoping Opportunities	16%	55 min
Inputs, Outputs and Contracts	16%	55 min
Review Points and Human Oversight	16%	55 min
Repeatability and Improvement	14%	45 min

Hands-on preparation checklist

Do these in a real ChatGPT account. Reading about repeatability teaches you nothing about repeatability.

Pick one recurring task you do weekly and score it on frequency, time cost, error tolerance and data sensitivity; decide out loud whether it is a good automation candidate.
Take a task you currently do in one long prompt and rewrite it as three named steps (for example extract → transform → format); note where the single prompt was hiding a failure.
Write a one-paragraph contract for each step: its inputs, required fields, output shape and acceptance criteria, and what happens on missing input.
Create a ChatGPT Project with custom instructions and 2–3 knowledge files for a recurring workflow, and run it twice to see whether output is stable.
Build a simple custom GPT for a task a colleague could run without you, then decide whether a Project would have been enough.
Use file uploads and data analysis on a spreadsheet, then Canvas to iterate on a document — and articulate which capability each task actually needed.
Place a review point in one workflow and define its trigger; then design a sampling check for a higher-volume version of the same task.
Write a runbook for one workflow so a colleague can run it end to end without asking you a question, and version the prompt with a date and a change note.