ProLehre | Medien und Didaktik Institute for LifeLong Learning Technische Universität München

Replikations-Paket · KI-Tutoren Bayern Replication Package · AI Tutors Bavaria

AI Guide Academy

Eine erprobte Schritt-für-Schritt-Vorlage, um generative KI über ein gefördertes Studierenden-Lehrenden-Modell in die Hochschullehre zu bringen — und an der eigenen Institution nachzubauen. A proven, step-by-step blueprint for bringing generative AI into university teaching through a funded student–instructor model — and replicating it at your own institution.

1

Das ProjektThe Project

Worum es geht und was es bewirkt hatWhat it is and what it achieved

Die AI Guide Academy bildet Studierende zu „AI Guides“ aus — Co-Piloten, die Lehrende dabei unterstützen, KI sinnvoll in ihre Lehre zu integrieren. An der TUM haben wir 100+ Tandems aus Lehrenden und Studierenden begleitet — die Schlagzahl liegt bei rund 25 Guides pro Semester. Zusätzlich haben wir in mehreren Workshops 500+ Lehrende aus ganz Bayern erreicht. The AI Guide Academy trains students to become “AI Guides” — co-pilots who help instructors meaningfully integrate AI into their teaching. At TUM we supported 100+ tandems of instructors and students — a steady cadence of about 25 guides per semester. On top of that, several workshops reached 500+ instructors from across Bavaria.

Problem & LösungProblem & Solution

Das ProblemThe Problem

Lehrenden fehlen Zeit, technisches Know-how und sichere Leitlinien, um KI-Werkzeuge wirksam in ihre Lehre einzubinden.Faculty lack the time, technical bandwidth and safety guidelines to effectively integrate AI tools into their teaching.

Das Co-Piloten-ModellThe Co-Pilot Model

Ausgebildete studentische Assistenzen wirken als Tech-Scouts, Tool-Entwickler:innen und Trainer:innen direkt im Fachbereich.Trained student assistants act as tech scouts, custom tool developers and departmental trainers — embedded directly in the course.

⏱️

Enorme ZeitersparnisMassive Time Savings

Guides übersetzten ganze Kurse ins Englische, erzeugten Code-Bausteine und bauten Datenbanken für jahrelange Altklausuren.Guides translated entire courses into English, generated code components and built databases for years of past-exam questions.

🛡️

Risiken abgefedertMitigated Risk

Sie navigierten den „Tool-Dschungel“, erkannten früh, wo kostenlose Tools versagen, und verhinderten Abhängigkeit von unskalierbarer Software.They navigated the “tool jungle”, spotted where free tools failed and prevented reliance on unscalable software.

🔧

Eigene InfrastrukturCustom Infrastructure

Statt generischer LLMs entstanden fachspezifische Tutor-Chatbots und automatisierte Feedback-Schleifen für Programmierübungen.Instead of generic LLMs, they built domain-specific tutor chatbots and automated feedback loops for programming exercises.

🏫

Train-the-TrainerTraining the Trainers

Guides schulten ganze Lehrstühle zu KI und schufen „lebende Dokumente“, damit das Wissen nach Projektende im Fachbereich bleibt.Guides trained entire chairs on AI and created “living documents” so the knowledge stayed in the department after the project ended.

2

Die wichtigsten EntscheidungenThe Key Decisions

Diese Weichen musst du vor dem Start stellenThe choices to make before you start

Replizieren ist kein Schema F. Diese fünf Entscheidungen prägen Budget, Aufwand und Form deines Programms — und sie hängen voneinander ab. Entscheidung A ist die Wurzel: Sie schaltet die Vertrags­frage (D) frei oder aus. Replication is not one-size-fits-all. These five decisions shape your budget, effort and program design — and they depend on one another. Decision A is the root: it switches the contract question (D) on or off.

A

Modell: bezahlt oder freiwillig?Model: funded or voluntary?

Die Grundsatzentscheidung. Sie bestimmt Budget, Rekrutierung und Anreizstruktur.The foundational choice. It determines budget, recruitment and incentives.

💶 Geförderter Track💶 Funded track

Guides erhalten Vergütung und werden formal einem/einer Lehrenden zugeordnet. Braucht Finanzierung und eine strukturierte Rekrutierungs-Pipeline.Guides receive a stipend and are formally paired with an instructor. Requires funding and a structured recruitment pipeline.

🎓 Freiwillige Fortbildung🎓 Voluntary training

Studierende absolvieren das Training freiwillig — ohne Förderung. Leichtere Infrastruktur, aber stärker auf intrinsische Motivation angewiesen.Students complete the training voluntarily — without funding. Lighter infrastructure, but relies on intrinsic motivation.

⚖️ Wenn du die freiwillige Fortbildung wählst, entfällt Entscheidung D (Vertrag & Vergütung) komplett.⚖️ If you choose the voluntary track, Decision D (contract & compensation) drops out entirely.
B

Institutionelle VerankerungInstitutional anchor

Wo sitzt das Programm formal? Beide Wege funktionieren — jeder spielt seine eigenen Stärken aus.Where does the program formally sit? Both routes work — each plays to its own strengths.

🏛️ Zentrale Einrichtung🏛️ Central unit

Fakultätsübergreifende Reichweite, vorhandene Verwaltung für Verträge & Förderung, hohe Glaubwürdigkeit und Nachhaltigkeit.Cross-faculty reach, existing admin for contracts & funding, strong credibility and sustainability.

🎯 Fakultät / Lehrstuhl🎯 Faculty / chair

Schneller Start, kurze Wege, hohe Fachnähe und volle Gestaltungsfreiheit im eigenen Bereich.Fast start, short paths, close disciplinary fit and full design freedom within your own unit.

C

Skalierung & TeamScale & team

Lege Zielzahlen fest, bevor du Infrastruktur aufbaust. Das Team skaliert mit der Größe — von einer Person in Teilzeit bis zu vier Rollen bei vollem Umfang.Set target numbers before building infrastructure. The team scales with size — from one part-time person up to four roles at full scale.

~25
Guides pro Semester (Schlagzahl)guides per semester (cadence)
100+
begleitete Tandemstandems supported

Faustregel: 1 Guide ↔ 1 Lehrende:r (ein Tandem). Vollausbau-Team: Programm-Lead ×1, Projektmanagement ×1, Werkstudierende ×2 — bei kleinem Umfang auch von einer Vollzeitkraft stemmbar.Rule of thumb: 1 guide ↔ 1 instructor (one tandem). Full-scale team: program lead ×1, project manager ×1, working students ×2 — at small scale a single full-time person suffices.

D

Vergütung & VertragCompensation & contract

nur bei bezahltem Trackfunded track only

An der TUM: HiWi-Vertrag über 6 Monate à 8 h/Woche, finanziert aus dem zentralen Budget per Kostenstellen­umbuchung („Finanzierung studentischer Hilfskräfte“) — der Lehrstuhl hat keine Kosten.At TUM: a student-assistant contract for 6 months at 8 h/week, funded from the central budget via a cost-transfer document — the chair incurs no costs.

Alternativen für andere Hochschulen: einmaliges Stipendium (Pauschale), Anrechnung als Leistungspunkte/Credits oder der reguläre interne HiWi-Prozess. An eigene Semester- bzw. Trimesterlängen anpassen.Alternatives for other institutions: a one-time lump-sum stipend, academic course credits, or your standard internal student-assistant workflow. Adapt to your own semester/trimester length.

E

Werkzeuge als BasisTools as your foundation

Hier geht es nicht um Technik um der Technik willen. Vier einfache Werkzeuge tragen die Kommunikation und Koordination im Programm — zwischen Guides, Lehrenden und Verwaltung. Nutze ruhig, was ihr schon habt; die Spalte rechts zeigt Alternativen inkl. Open Source.This isn't about technology for its own sake. Four simple tools carry the communication and coordination in the program — between guides, instructors and administration. Use whatever you already have; the right column shows alternatives, including open source.

📝 Formulare📝 Forms

Um Bewerbungen einzusammeln und später Feedback abzufragen.To collect applications and, later, gather feedback.

Microsoft FormsGoogle FormsLimeSurvey (OSS)Nextcloud Forms (OSS)
🗃️ Übersicht & Zuordnung🗃️ Overview & pairing

Um den Überblick zu behalten, wer dabei ist, und Guides mit Lehrenden zusammenzubringen.To keep track of who's taking part and pair guides with instructors.

Teams + ExcelAirtableNocoDB (OSS)Baserow (OSS)
🎓 Lernplattform🎓 Learning platform

Um das Training bereitzustellen, das die Guides in ihrem Tempo durcharbeiten.To host the training that guides work through at their own pace.

Moodle (OSS)ILIAS (OSS)CanvasOpenOlat (OSS)
🤖 Zugang zu KI-Tools🤖 Access to AI tools

Damit die Guides tatsächlich mit KI arbeiten können — kommerziell oder datenschutzfreundlich selbst gehostet.So guides can actually work with AI — commercial or self-hosted for data privacy.

ChatGPTClaudeGeminiCopilotUni-eigenes KI-ToolUniversity-hosted AI
3

Dokumente & VorlagenDocuments & templates

Material zum direkten WiederverwendenMaterial you can reuse directly

Lade die zentralen Artefakte herunter und passe sie an. Das Moodle-Backup spielst du direkt in deine Instanz ein; die Lernziele kannst du als Curriculum-Grundlage übernehmen.Download the core artefacts and adapt them. Import the Moodle backup straight into your instance; reuse the learning goals as your curriculum baseline.

🎓

Der TrainingskursThe training course

Der vollständige AI-Guide-Trainingskurs — mit Moodle importierbar oder ohne Moodle direkt im Browser lesbar.The complete AI Guide training course — importable with Moodle or readable straight in the browser without it.

Die Browser-Ansicht enthält alle vier Module mit den kompletten Kapiteltexten und Bildern. Interaktive H5P-Elemente, Videos und Feedback-Aktivitäten sind nur im Moodle-Backup enthalten (in der Ansicht als Hinweis markiert).The browser view contains all four modules with the full chapter texts and images. Interactive H5P elements, videos and feedback activities live only in the Moodle backup (flagged as notes in the view).

🎯

Lernziele-CurriculumLearning-goals curriculum

Das komplette Curriculum in drei Modulen mit insgesamt 25 Lernzielen. Originaldatei als Excel.The full curriculum across three modules with 25 learning goals in total. Original file as Excel.

Wie der Kurs aufgebaut istHow the course is built

Struktur: Jeder Abschnitt ist ein Prüfungsthema. Inhalte stecken in „Büchern“; jedes Buch enthält ein Lernziel und hat eine zugehörige Feedback-Aktivität.Structure: each section is an exam topic. Content lives in “books”; every book holds one learning goal and has a matching feedback activity.

Video-Aufbau: Teaser → Warum wichtig? → Agenda → „Meat“ (Inhalt in Video- und Textform + Aufgaben) → Zusammenfassung.Video format: teaser → why it matters → agenda → the “meat” (content in video and text + tasks) → summary.

Feedback: zwei 7-stufige Likert-Fragen (Gefiel dir der Inhalt? / Wärst du enttäuscht ohne ihn?) plus eine offene Frage.Feedback: two 7-point Likert questions (Did you like the content? / Would you be disappointed without it?) plus one open question.

Modul 1 · KI-GrundlagenModule 1 · AI Basics10
Code ThemaTopic StufeLevel LernzielLearning goal
AIB-00CapstoneAnalysierenAnalyseVerschiedene KI-Tools bewerten und entscheiden, welches eingesetzt wirdEvaluate different AI tools and decide which one to use
AIB-01LLMsVerstehenUnderstandDie Unterschiede zwischen KI, LLM und ML erklärenExplain the differences between AI, LLM and ML
AIB-02LLMsVerstehenUnderstandErklären, wie LLMs funktionieren und wo ihre Grenzen liegenExplain how LLMs work and their limitations
AIB-03PromptingPromptVerstehenUnderstandVerstehen, wie man KI effizient einsetztUnderstand how to treat AI efficiently
AIB-04PromptingPromptAnwendenApplyPrompts für unterschiedliche Probleme erstellenCreate prompts for different problems
AIB-05EvaluationEvalAnalysierenAnalyseLLM-Leistung anhand von Benchmarks analysierenAnalyse LLM performance against benchmarks
AIB-06EvaluationEvalAnalysierenAnalyseBeurteilen, ob KI ein Problem lösen können sollteJudge whether AI should be able to solve a problem
AIB-07EvaluationEvalAnalysierenAnalyseAnalysieren, ob ein Tool seinen versprochenen Nutzen liefertAnalyse whether a tool delivers its promised value
AIB-08GrenzenLimitsAnalysierenAnalyseEntscheiden, ob ein geplanter KI-Einsatz rechtlich zulässig istDecide whether a proposed use of AI is legal
AIB-09GrenzenLimitsBewertenEvaluateZu moralischen Dilemmata rund um KI Stellung beziehenTake a stance on moral dilemmas raised by AI
Modul 2 · DidaktikModule 2 · Didactics8
Code ThemaTopic StufeLevel LernzielLearning goal
DID-0CapstoneErstellenCreateEinen Workshop von Grund auf konzipierenDesign a workshop from scratch
DID-1LehrenTeachBewertenEvaluateEinen Workshop-Plan mit Constructive Alignment verbessernImprove a workshop plan with constructive alignment
DID-2LehrenTeachAnwendenApplyMit Blooms Taxonomie sinnvolle Lernziele formulierenUse Bloom's Taxonomy to express meaningful learning goals
DID-3LehrenTeachVerstehenUnderstandKognitive Aktivierung als Prädiktor für Lernerfolg verstehenUnderstand cognitive activation as a predictor of learning success
DID-4LehrenTeachAnwendenApplyEine Sitzung nach dem ARIVA-Schema strukturierenStructure a session according to ARIVA
DID-5LehrenTeachAnwendenApplyEine PowerPoint-Folie sinnvoll überarbeitenTransform a PowerPoint slide for the better
DID-6PrüfenTestVerstehenUnderstandDie grundlegenden Gütekriterien für Prüfungen benennen und verstehenName and understand the basic quality criteria for tests
DID-7PrüfenTestAnwendenApplyEine Prüfung erstellen, die das Erreichen der Lernziele überprüftCreate a test to check if learning goals are met
Modul 3 · KI in der LehreModule 3 · AI in Education7
Code ThemaTopic StufeLevel LernzielLearning goal
AIE-0CapstoneErstellenCreateWorkflows und Prompts erstellen, um die Kursvorbereitung teilweise zu automatisierenCreate workflows and prompts to semi-automate course prep
AIE-1LernzieleLearning goalsErstellenCreateLernziele aus mehreren Quellen erstellenCreate learning goals from multiple sources
AIE-2MaterialErstellenCreateFundierte Dokumente zu einem konkreten Thema erstellenCreate solid documents around a specific topic
AIE-3MaterialErstellenCreateFolien erstellen, die Lehrende durch eine Sitzung führenCreate slides that guide a lecturer through a session
AIE-4MaterialErstellenCreateAktivitäten erstellen, die Studierende zum Lernziel führenCreate activities that help students reach the goals
AIE-5TestsErstellenCreatePrüfungen erstellen, die die Lernziele abbildenCreate tests that reflect the learning goals
AIE-6TestsErstellenCreatePrüfungen erstellen, die KI (noch) nicht lösen kannCreate tests that AI can't beat (yet)
📄

Operative VorlagenOperational templates

Bewerbungsformulare, Zusage-E-Mails, das „AI Guide Info“-Formular und die Kostenstellen-Umbuchung sind in Phase 4 (Fahrplan) bei den jeweiligen Schritten als Klapptext hinterlegt — bereit zum Kopieren und Anpassen.Application forms, acceptance emails, the “AI Guide Info” form and the cost-transfer document are embedded as collapsible text within the matching steps in Phase 4 (Roadmap) — ready to copy and adapt.

4

Dein FahrplanYour roadmap

Beantworte 5 Fragen → maßgeschneiderte ChecklisteAnswer 5 questions → tailored checklist

Beantworte die fünf Fragen — passend zu den Entscheidungen aus Phase 2 — und wir erstellen daraus eine auf dich zugeschnittene Checkliste. Darunter findest du den vollständigen, ein- und ausklappbaren Ablauf.Answer the five questions — mirroring the decisions from Phase 2 — and we'll build a checklist tailored to you. Below it you'll find the full, expandable step-by-step plan.

A · Bietet ihr das Programm bezahlt oder als freiwillige Fortbildung an?Will you offer the program funded or as voluntary training?
B · Wo verankert ihr das Programm?Where will you anchor it?
C · Wie viele Guides im ersten Durchlauf?How many guides in the first cohort?
D · Wie lang ist euer Durchlauf?How long is your term?
E · Welches Tool-Ökosystem nutzt ihr schon?Which tool ecosystem do you already use?

Der vollständige AblaufThe full step-by-step plan

1 Planung & SetupPlanning & Setup~3 Monate vor Kursbeginn~3 months before course start

Lege Förderung, Skala, Zeitplan, Verankerung, Team und Tech-Stack fest (siehe Phase 2). Empfohlener Host: eine zentrale Lehreinrichtung.Lock down funding, scale, timeline, anchor, team and tech stack (see Phase 2). Recommended host: a central teaching center.

Zeitplan (relativ zum Kursstart)Timeline (relative to course start)

  • Werbung — 3 Monate vorher: offener Call + Lehrende einladenAdvertisement — 3 months before: open call + invite instructors
  • Rekrutierung — ab 2 Wochen nach Call: screenen & bestätigenRecruitment — from 2 weeks after call: screen & confirm
  • Training — zum Kursstart: Onboarding der GuidesTraining — at course start: onboard guides
  • Live-Sessions — Monate 1–2Live sessions — months 1–2
  • Austauschtreffen — Monate 3–5Exchange meetings — months 3–5
  • BarCamp — Monat 5/6 · Abschluss — Monat 6BarCamp — month 5/6 · Wrap-up — month 6
2 Marketing & RekrutierungMarketing & RecruitmentBewerbung, Auswahl, MatchingApplication, selection, matching

BewerbungsformulareApplication forms

Je ein Formular für Lehrende und für Guide-Kandidat:innen. Erhobene Felder:One form for instructors, one for guide candidates. Fields collected:

  • Lehrende: Name, E-Mail, Fakultät, „Wie auf uns aufmerksam geworden?“, Freitext (3.000 Z.): Wie könnte KI deine Lehre verbessern, was hast du schon probiert, wie würdest du den Guide einsetzen?Instructors: name, email, school, “how did you hear about us?”, free text (3,000 chars): how could AI benefit your teaching, what have you tried, how would you use your guide?
  • Studierende: Name, E-Mail, Fakultät, Quelle, Deutsch-/Englisch-Niveau, Freitext: Erfahrungen mit KI & Motivation als GuideStudents: name, email, school, source, German/English proficiency, free text: AI experience & motivation as a guide

Faire Verteilung (nur bei zentraler Verankerung)Fair allocation (central anchor only)

Förderstellen werden nicht nach „first come“ vergeben, sondern proportional zur Studierendenzahl je Fakultät:Funded slots aren't awarded first-come — they're allocated proportionally to each school's share of students:

  • Studierende der Fakultät ÷ Studierende gesamt × Förderstellen = Kontingent der FakultätSchool's students ÷ total students × funded slots = school's allocation
  • Beispiel: 25 Stellen, CIT = 29 % der Studierenden → 7 StellenExample: 25 slots, CIT = 29 % of students → 7 slots

Matching: 3 WegeMatching: 3 routes

  • Bekannte:n Studierende:n direkt benennenNominate a known student directly
  • Mit Stellen-Vorlage öffentlich ausschreibenAdvertise publicly using the job template
  • Aus dem Kandidat:innen-Pool (Datenbank) wählenPick from the candidate pool (database)

Vorlage: Zusage-E-Mail an LehrendeTemplate: instructor acceptance email

3 HR & FinanzenHR & Financenur bei bezahltem Trackfunded track only

Der Finanzierungs-LoopThe funding loop

  • Auslöser: „AI Guide Info“-Formular (Lehrende + Studierende: Namen, E-Mails, Startdatum, Wochenstunden, Notizen)Trigger: “AI Guide Info” form (instructor + student names, emails, start date, weekly hours, notes)
  • Kostenumbuchung: „Finanzierung studentischer Hilfskräfte“ — autorisiert 8 h/Woche über 6 Monate aus dem ZentralbudgetCost transfer: the “Finanzierung studentischer Hilfskräfte” document — authorises 8 h/week for 6 months from the central budget
  • HR-Übergabe: Lehrende:r reicht das Dokument bei der HR des Lehrstuhls ein → offizieller HiWi-VertragHR handover: instructor submits the document to the chair's HR → official student-assistant contract

Onboarding & Tool-ZugangOnboarding & tool access

  • Willkommens-E-Mail mit Moodle-Zugang + GruppenschlüsselWelcome email with Moodle access + group key
  • Onboarding-Umfrage: Vertragsstart, Live-Training-Termin, bevorzugtes LLMOnboarding poll: contract start, live-training date, preferred LLM
  • LLM-Zugang bereitstellen (kommerziell oder uni-gehostet)Provision LLM access (commercial or university-hosted)

Alternativen ohne HiWi-Vertrag: einmaliges Stipendium, Leistungspunkte oder interner Hilfskraft-Prozess.Alternatives without a contract: one-time stipend, course credits, or your internal assistant process.

4 TrainingTrainingSelf-Paced auf Moodle + CapstoneSelf-paced on Moodle + capstone

Die Guides arbeiten das Training selbstständig durch und schließen je ein Capstone ab. Den fertigen Kurs (drei Module, 25 Lernziele) musst du nicht neu bauen — spiele einfach unser Moodle-Backup ein.Guides work through the training self-paced and complete one capstone each. You don't need to rebuild the course (three modules, 25 learning goals) — just import our Moodle backup.

5 Live-DurchführungLive ExecutionWorkshops & AustauschtreffenWorkshops & exchange meetings
  • Kick-off-Workshop planen & durchführenSchedule & run the kick-off workshop
  • Regelmäßige Guide-↔-Lehrende-SessionsRegular guide-↔-instructor sessions
  • Hochschulübergreifende AustauschtreffenInter-university exchange meetings
  • Feedback sammeln & für die nächste Kohorte iterierenCollect feedback & iterate for the next cohort
6 Abschluss & BarCampWrap-up & BarCampMonat 5/6Month 5/6
  • Ganztägiges BarCamp mit Peer-Präsentationen & KI-GamificationFull-day BarCamp with peer presentations & an AI gamification challenge
  • Abschlussberichte der Guides einsammelnCollect final project reports from guides
  • Verträge & LLM-Lizenzen sauber beendenClose out contracts & LLM licenses
  • Zertifikate verteilen & Feedback erhebenDistribute certificates & gather feedback