Paris School of EntrepreneurshipAI in Practice — Level 2 · Lucas Zanelatto
Session 1 of 6
How AI Actually Works0%
Session 1 —
How AI Actually Works
Wednesday 2 September · 09:00–13:00 · Servcorp Paris Kléber
Enter your access code to begin.
Your code was in your Student Portal email.
1
Session 1 · Wednesday 2 September · 09:00–13:00 · Servcorp Paris Kléber
How AI Actually Works — In-person
Session Brief
Before we begin — what did you bring today?
Three questions to open the session. The more specific your answers, the more targeted the three days become.
Opening · Q01
Your label, your artists, your AI video idea — in two sentences.
Your coach has read your Discovery Questionnaire. This is how you say it out loud for the first time.
Opening · Q02
Where have you tried AI so far — and what was the most frustrating result?
The frustration reveals the exact gap we close in Block 1.
Opening · Q03
What would you need to see produced by AI today to feel this morning was worth it?
Write it down. Your coach comes back to it at 13:00.
01Build the mental model of how AI actually works — tokens, context, hallucination — anchored to your label's use cases.
02Master the RCTF prompting framework and apply it to 3 real label tasks before the break.
03Map the AI tool landscape to your specific workflow — which tools for which tasks, starting this week.
Session 1 Coaching Deck
Full Session
Session 1 Slide Deck
Your coaching slides for the full session. Arrow keys to navigate. Right-click disabled.
Block 1 Complete · Thinking Check 1 of 2
AI Foundations
Block 1 — Thinking Check
Three questions on AI fundamentals applied to your label. Complete during the break — your coach reviews while you rest.
Check 1 · Question 01
When AI produces a confident but incorrect answer about a competitor's streaming numbers — what is the underlying reason?
Apply what you just learned — not a general definition.
AI searched the internet and found wrong data
AI predicted the most statistically likely completion of the text — not the most accurate one
The core mechanic
AI is deliberately simplifying to give a faster answer
The model was not trained on music industry data
Check 1 · Question 02
Which of the three AI limitations is most immediately dangerous for your music label — and why?
Hallucination — AI fabricates artist bios, chart data, booking contacts, contract terms
Context blindness — AI forgets label context between sessions, producing generic outputs
Recency cutoff — AI has no knowledge of recent releases, streaming rule changes, current trends
All three are equally dangerous for different parts of the business — be specific about which part
Strongest answer
Check 1 · Question 03
What will you do differently when using AI for your label after this block?
The most specific answer is the most useful one — for you and your coach.
0 / 400
Thinking Check Score
—
Deliverable 1A · Prompt Library Starter
Block 2 — Build Your First RCTF Prompt
Build your first RCTF Prompt
Assemble a real prompt for your label using the four-component framework. This becomes the first entry in your Prompt Library — one of your Friday capstone deliverables.
RCTF Prompt Builder — Live
Your Label Prompt
Fill in each component. Your complete prompt assembles below in real time.
R — Role
C — Context
T — Task
F — Format
Your assembled prompt
Your assembled prompt will appear here as you fill in each component above.
Block 2 Complete · Thinking Check 2 of 2
Prompt Engineering
Block 2 — Thinking Check
Four questions applying the RCTF framework to your label's real use cases.
Check 2 · Question 01
Which component of RCTF makes the biggest difference to your outputs — and why?
Role — without it, everything defaults to generic mode
Context — AI knows nothing about my label unless I provide it every time
Task — vague tasks produce vague outputs; specificity forces a specific answer
Format — without it, I always have to reformat the output manually
Check 2 · Question 02
Name two specific label tasks where a well-built RCTF prompt would save you more than 30 minutes per week.
Be specific — task name, approximate time saved. Vague answers here mean missed opportunities.
0 / 500
Check 2 · Question 03
You run a prompt and the output is mediocre. Using RCTF, what do you check first?
Rephrase the whole thing and try again
Diagnose which component is missing or weak — role, context, task, or format
Operator approach
Switch to a different AI tool
Check 2 · Question 04
Write a chain-of-thought prompt for one real label challenge — ask AI to show its reasoning step by step.
Chain-of-thought: ask AI to 'think through this step by step' before the final answer. It transforms surface-level outputs into genuine analysis.
0 / 600
Thinking Check Score
—
Deliverable 1B · AI Landscape Map
Block 3 — Match Tools to Tasks
Your AI Landscape Map
Match the right AI tool to each label task. Fill in at least 4 of the 6 task areas — this becomes your personal AI operating guide.
1
Task Area 1
Artist research & competitive intelligence
Perplexity for cited current data. Claude for deep analysis of longer reports.
2
Task Area 2
Writing — artist bios, press releases, pitches
Role: music industry analyst. Context: artist name, genre, key achievements, target audience.
3
Task Area 3
AI video creation — music videos & content
Runway Gen-4, Kling AI, Pika — each has different strengths for music video style and length.
4
Task Area 4
Social & marketing content
5
Task Area 5
Strategic decisions & analysis
Chain-of-thought prompting is most valuable here — ask AI to reason through assumptions before recommending.
6
Task Area 6
Client work — other businesses
How do you separate your label tools from your client tools?
Session 1 Complete.
You've built the mental model, mastered the RCTF framework, and mapped AI to your label workflow. Session 2 starts at 14:00 — same room.
RCTF Prompt Library Starter
Your first structured prompt — the foundation of Friday's complete Prompt Library.
AI Landscape Map
Six task areas matched to the right AI tool and approach — your personal AI operating guide.
AI Mental Model — confirmed
Tokens, context, hallucination, recency cutoff — anchored to your label's specific risks.
Slides
← → keys · right-click disabled
1/1