EN Language: English

Build your sales tool

A course where you build your own sales tool with AI and a person signs off your exam.

CHF 199 one-off · 30 hours · taught in German

Swiss Academy for Digital Skills, Zürich

Buy the course

Change allowed

Status rules from Mission 3: a lead may only move along certain paths. The checking service tests exactly that, in criteria C02-M3-06 and C02-M3-07.

Limmat Solar AGLuca Zimmermann · L-0005

New to Contacted: allowed. Criterion C02-M3-06.

Email
luca.zimmermann@limmat.example
Follow-up
05.03.2026

Only the server counts

On your own machine, anything goes. What counts is only what the checking service confirms with the real tests in an isolated environment.

A ZIP with missions, tests and instructions for your agent. Claude Code or Codex reads CLAUDE.md and sets everything up with you.

c02-starter-kit.zip
  • CLAUDE.mdAnleitung für deinen Agenten
  • briefing.mdDer Auftrag des Kunden
  • KI-PROTOKOLL.mdWas die KI beigetragen hat
  • missions/Acht Missionen und die Prüfung
  • src/Die App, mit TODO-Stellen
  • tests/Die Kriterien jeder Mission
  • bic.config.jsonVerbindung zum Prüfdienst

You describe, the agent writes, you read it through. You record what the AI contributed in the AI log. Hints come in up to three levels and never give away the solution.

KI-PROTOKOLL.md

Mission 2, Lead-Liste

Vorschlag der KI
Tabelle mit einer Zeile pro Lead und onClick auf der Zeile.
Übernommen oder geändert
Übernommen, aber Button in der Zeile ergänzt, damit es mit Tastatur geht.
Geprüft durch
check 2 grün und manuell mit der Tab-Taste ausprobiert.

One command shows which criteria pass and which do not. Two deliberate bugs in the starter are waiting to be found.

$ npm run bic -- check 3
Mission 3 (c02-v0.1.0): NOCH NICHT
[ok] C02-M3-01 speichert einen Lead und liefert ihn in list() als Kopie zurueck
[ok] C02-M3-02 behaelt Daten, wenn das Repository mit demselben Speicher neu erstellt wird
[ok] C02-M3-03 ueberschreibt bei gleicher id, findet per get() und entfernt per remove()
[  ] C02-M3-04 validateLead lehnt fehlende Firma und ungueltige E-Mail mit Feldmeldungen ab
[  ] C02-M3-05 loadSeedLeads sortiert genau die sechs fehlerhaften Beispielzeilen aus
       AssertionError: expected [ …(50) ] to have a length of 44 but got 50
[ok] C02-M3-06 erlaubt die Statuswechsel aus der Tabelle
[ok] C02-M3-07 verweigert unerlaubte Wechsel und veraendert das Original nie
[ok] C02-M3-08 filtert nach Status und gibt ein neues Array zurueck
[ok] C02-M3-09 filtert ueberfaellige Follow-ups relativ zum Stichtag
7 von 9 Kriterien erfuellt in 3.8 s
Beispiel: Mission 3, bevor validateLead fertig ist.

The checking service swaps your tests for the real ones, runs them offline in an isolated container and writes criteria, XP and achievements to your progress.

Lernstandserverseitig geprüft

Mission 3, Daten und Regeln beherrschen

9 von 9 Kriterien erfüllt

  • A04 Daten im Griff+100 XP
  • A05 Regeln funktionieren+100 XP

Mission 4 freigeschaltet

Written and online: you answer six questions about your own project whenever it suits you. An AI pre-assesses against the published rubric, then a human examiner reviews and releases the result.

Prüfungsrubrikbestanden ab 75
BereichPunkteMinimum
Funktion und Nutzen25
Verständnis2515
Tests und Fehlersuche20
Zugriff und Secrets2014
Übergabe und Bedienbarkeit10
Black and white photo: about twenty people standing and sitting in front of the control desk and cabinets of ERMETH, with the word ERMETH on the wall.

Build, don't just watch

There are no videos to watch along with. You build a tool that is yours at the end, and you learn what still matters when the agent writes the code.

Zürich, 1963: group photo at ERMETH, the first computer built at ETH Zürich.

Course

7 missions, a transfer task and an exam, 30 hours. Every mission ends with criteria that the server checks.

  1. Set up your tools1 h

    By the end, the app runs in your browser, the tests start, and your AI agent knows the course.

    Read the mission
  2. Understand the brief2 h

    A clear brief with testable limits before you write a line of code. If you cannot describe the problem, you cannot give the AI a good brief either.

    Read the mission
  3. Build the interface4 h

    The sample data becomes a usable interface: a list of all leads and a detail view with a way back. You learn how React turns data into screens and what a data-testid is.

    Read the mission
  4. Master data and rules5 h

    The app keeps data and states consistent: saves survive a reload, invalid input is rejected, status changes follow fixed rules, filters work. That is the core of any business application.

    Read the mission
  5. Prepare the next steps5 h

    Your app talks to an API and stays clear even when the API returns nothing, crashes or hangs. That is what separates a work sample from a prototype.

    Read the mission
  6. Find bugs and check access4 h

    You find two planted bugs, fix them so that tests prove it, restore a working version and check that access and secrets are in order.

    Read the mission
  7. Present and hand over4 h

    Someone else can start your project, understand it and know its limits. And you can explain what you built. This prepares you for the written exam.

    Read the mission
  8. Transfer3 h

    At work, requirements arrive without guides, test lists or hints. The transfer checks whether you and your agent can tackle a new task so it is testable, clean and explainable. No visible tests, no hints.

    Read the mission
  9. Exam

    Written and online: you answer six questions about your own project whenever it suits you. An AI pre-assesses against the published rubric, then a human examiner reviews and releases the result.

    Read how it works

Read all missionsGet the starter kit

What you build: a lead list with sample companies, rules, filters and an API.

Verifiable, not claimed

Every certificate carries a QR code and a link. Anyone who opens it sees the course, version, skills, date and status, with no login and no person search. A revocation is shown there too.

Private academy certificate, not state accredited.

Zertifikat · Muster

AI App Builder: Eigenes Sales-Tool

Vorname Nachname

  • Eine Arbeitsaufgabe in überprüfbare Teilschritte zerlegen und als Akzeptanzfälle beschreiben.
  • Eine einfache Anwendung mit einem KI-Agenten erstellen und nachvollziehbar verändern.
  • Fehlfälle, Datenprüfung, Zugriff und Geheimnisse testen und absichern.
  • Die eigene Lösung erklären, ihre Grenzen benennen und auf eine neue Anforderung übertragen.
Kursversion
c02-v0.1.0
Status
gültig
Prüflink
app.builtinclass.com/verify/…

Pay once, no subscription

Course C02, AI App Builder

  • 30 hours, 7 missions and a transfer task
  • Checks on the server, hints in up to three levels
  • One written exam, released by a person
  • 3 months of access, can be extended

CHF 199

Buy the course

Second exam, CHF 149. With a pre-exam report and the right to appeal, for learners with active access.

Companies. Team seats on invoice, one cohort with a shared exam window. Send a request

Prices in CHF including VAT, payment by card.

Questions

Is the course in English?

No. The course, its materials, the platform and the exam are in German only. If English is your main language, you need solid German reading skills, and the exam is held in German.

Do I need programming skills?

No. You need a subscription to Claude Code or Codex, Node.js and a willingness to read carefully. The agent writes the code. You break the task down, check, decide and explain.

Can I tamper with the tests locally?

Locally yes, on the server no. The checking service replaces the test files with the canonical version and runs them offline in an isolated container. Only that result counts.

How does the exam work?

You start the exam when you are ready and answer six written questions within 14 days: one per rubric area about your own project, plus a transfer question. No appointments, no conversation. An AI (Claude by Anthropic) pre-assesses your answers against the published rubric. An examiner then reviews them manually and releases the result. Marking follows the rubric with a minimum per area.

What if I fail?

You get the reasons based on the rubric, can appeal within 14 days and can book a second exam.

Who owns my code?

You do. We only receive the work you submit, check it in isolation and delete it on request. Servers in the EU. For the exam, your answers and excerpts of your project go to Anthropic (USA) for the AI pre-assessment.

What is the certificate worth?

A private academy certificate with a publicly verifiable link, not state accredited. It shows what you built and explained in the exam: course, version, four skills, date, status. It confirms the work submitted under your account and released by an examiner.