FR EN
2026 🚀
Verrier × Alexis Pineau · 2026

Artificial
Intelligence

Understand it, use it, and save time every day

Scroll
01

What comes to mind when you hear “AI”?

🤖
Robots
🎬
Science fiction
💀
Terminator
🧠
Digital brain
😟
Worried?
Relax. It is far simpler than that.
02

Today, AI is mostly used as
an assistant.

🧠 An assistant 📚 Someone who has read everything ⚡ Very fast
Answer!
03

Where does it come from?

💡
💡
1950
Researchers ask the question
“Could a machine one day think, or solve problems the way a human does?” — Alan Turing, 1950. The idea of artificial intelligence is born.
millions of games
1950–90
Learning through repetition
It doesn’t follow rules — it plays millions of games. Rewarded when it wins a piece. Penalised when it loses one. It adjusts, repeats, improves.
Reinforcement Learning
🏆
Deep Blue
VS
Kasparov
🤖 > 👑
1997
Deep Blue beats the world champion
The machine beats Garry Kasparov, the best chess player in the world. It is a shock. AI starts to be taken seriously. But it can still only do one thing.
Turning point
📈
power × data
2010s
The quiet explosion
More data, more powerful computers, new methods. AI starts recognising faces, understanding voices, reading text. The revolution is building.
💬
What exactly is it?
I’m ChatGPT, I can help you write, summarise, analyse…
Impressive 🤯
2022
ChatGPT arrives. Everything changes.
For the first time, an AI can write, summarise, analyse, code and translate — and anyone can use it. For free. From a browser. In five minutes.
This is now
The first time

In class, with a friend.
We had no idea what to expect.

We were in sixth form. ChatGPT had just launched. We typed the first thing that came to mind.

chat.openai.com
What is ChatGPT?
🤖
I am ChatGPT, a language model created by OpenAI. I’m designed to understand and generate text in response to your questions and requests…
— ChatGPT, November 2022
🧪
We tested the limits
Jokes, maths, forbidden topics… we were trying to break it.
📚
Then genuinely useful
Revision sheets, lecture summaries, help understanding a difficult concept.
🚀
It kept evolving
Each version stronger than the last. Today the pace is dizzying.
And now we use…
🤖
ChatGPT
The pioneer. Mass market. Office integrations, DALL·E, web browsing.
WHAT WE USE HERE
Claude
Deep reasoning. Very reliable on long documents. Ideal for complex tasks.
Hands-on exercise

Your turn. 🎯

Here is a real business email — dashed off, full of mistakes. Your mission: fix it with ChatGPT in 30 seconds.

🤖 Open ChatGPT
1
Copy this email
From: marc.dupont@verrier-bureau.fr
To: client@societex.fr
Subject: Office quote
Hello, following our phone conversation last week, i'm sending you the quote for the desk you asked for. there is 6 solid oak desk + the matching chairs. the total price it is 4 200€ excl. VAT. if you have any question, don't hesitate to contacting me. Regards Marc
2
Paste it into ChatGPT with this prompt
You are a professional assistant. Correct this sales email: improve the spelling, tone and structure. Keep the message short and professional. Here is the email: [PASTE THE EMAIL HERE]
In 30 seconds you have an email that is professional, error-free, and in the right tone.
That is what everyday AI looks like.
Exercise 2 — Catalogue analysis

A PDF catalogue. A table in 10 seconds.

ChatGPT can read a PDF and pull out exactly what you need — without you working through it page by page.

📥
Download the catalogue
CatalogueMBP2021.pdf
📎
Attach it to ChatGPT
Paperclip button → select the file
✍️
Send the prompt
Copy and paste from below
The prompt to use
You are an office furniture expert. Analyse this PDF catalogue and select the 10 best chairs. Present the result as a table with the following columns: - Model name - Indicative price - Type (task, executive, visitor…) - Strengths (2-3 words) - Best suited for Add a short recommendation below the table: which model you would choose for a 20-person open space, and why.
Result: a structured table plus a recommendation — without opening the PDF once.
What used to take you 20 minutes now takes 10 seconds.
Exercise 3 — AI vision

Show it an image.
It understands what it sees.

ChatGPT can analyse a photo, a screenshot or a scanned document — and answer questions about it as though it had read it.

🖥️
Screenshot
Competitor site, quote, incoming email
📸
Product photo
Desk, chair, layout to identify
🧾
Scanned document
Invoice, purchase order, floor plan
📊
Chart / table
Stats, results, readings
chat.openai.com
🖼️ competitor_site_screenshot.png
What do you see in this screenshot? Give me the competitor’s strengths and how we could do better.
🤖
I can see a product page with 3 selling points highlighted : fast delivery, 5-year warranty, online configurator. Their visible weakness: no lifestyle photography, everything is on a white background — you could stand out with real interior shots.
Try it with any image
Look at this image and tell me: 1. What you see (products, context, atmosphere) 2. The strengths and the weaknesses 3. What you would do differently to improve it Be precise and direct.
A photo taken on your phone. An instant analysis.
AI doesn’t only read text — it sees.
Exercise 4 — Real-time search

It can search the internet.
Live.

ChatGPT is no longer limited to its training data. Turn on web search and ask it any current question — it will search, read and summarise for you.

🔍
How to turn on search
In ChatGPT, click the 🌐 Search icon (or the small globe) before sending your message. ChatGPT shows the sources it consults in real time.
chat.openai.com 🌐 Search enabled
What are the current trends in office design in 2024? Summarise the 3 main trends.
🤖
🌐 Searching… Reading 4 sources…
According to several recent studies, the 3 main trends are:

1. The flexible hybrid office — modular spaces for alternating between on-site and remote work.
2. Wellbeing at the centre — natural light, acoustics, ergonomic furniture.
3. Sustainability — recycled materials, refurbished furniture, environmental certifications.

Sources: Steelcase Workplace Survey 2024, Leesman Index, IFMA France
Try this prompt (with search enabled)
No more spending 30 minutes on Google piecing articles together. It monitors the market for you.
News, competitors, market trends — in real time.
The only rule

There is no wrong way to do this.

AI doesn’t break. You can’t “use it wrong”. The only way to learn is to use it — and discover as you go what it can do.

🧪
Try anything
An odd idea? A silly question? Ask it. That is often where it gets interesting.
🔁
Start over if it misses
Not happy with the answer? Say so. Rephrase. It is a conversation, not a form.
🚀
Start today
No training needed. No need to be comfortable with tech. Just a first question.
Open ChatGPT and get started
04

The key: the prompt

The more precise you are, the better the answer.

Poor prompt
“Analyse this document.”
→ Vague result, unusable
Good prompt
“You are an accounting assistant. Here is a supplier quote as a PDF. Extract every line into an Excel table with: reference, description, quantity, unit price excl. VAT, total excl. VAT. Final format: a clean table, sorted by category.”
→ A clean table in 10 seconds

That is the difference between an assistant you brief badly and one that does exactly what you want.

05

What you need to know

⚠️
Hallucination
AI can invent facts with total confidence. Always verify anything that matters.
🔒
Data
Never paste passwords or sensitive customer data. Turn off history in the settings.
🌱
Environment
Training an AI consumes a lot. Using it day to day: roughly the same as a Google search.
✓ Done!
90
0%

That is what you need to know to use AI day to day.

But we are going further still. ↓

07

The real shift happening now

🎯
It tells complex tasks apart from simple ones and adjusts its level of effort.
🧩
It can reason step by step — breaking a problem down before answering.
📁
It can access YOUR files. That is the fundamental change.
📄
📊
📧
08
📁
📄
🧠

The Second Brain

Imagine an AI that knows your company inside out — its customers, its prices, its procedures. That is exactly what a Second Brain makes possible.

The analogy
Without a Second Brain
You ask ChatGPT to draft a proposal. It doesn’t know who your customer is, or your prices, or how you work. The answer is generic.
With a Second Brain
You give it access to your files. It knows your customers, your rates, your context. The answer is tailored.
What goes inside
📂 second-brain-verrier/
📋 grille-tarifs-2024.xlsx Your prices, your margins
👥 clients-fideles.csv Purchase history
📝 procedure-relance.md How you work
📦 catalogue-produits.pdf References, descriptions
✉️ modeles-mails.md Tone, phrasing, sign-offs
How it works
1️⃣
You create a folder
On your computer. Nothing in the cloud. Your data stays with you.
2️⃣
You put your files in it
Rates, procedures, customers — anything useful to your work.
3️⃣
The AI has access to it
Claude reads all of it before answering. Its replies become precise, contextual and actionable.
Example — what it looks like at Verrier
Draft a follow-up email for the town hall of X — they ordered desks 6 months ago and we haven’t heard from them since.
Read: clients-fideles.csv, modeles-mails.md, grille-tarifs-2024.xlsx

Subject: Following up on your order — and something new that may interest you

Dear Sir or Madam,
Following your order last June, I wanted to check in with you… [full email, personalised with your data]
🪨
Recommended tool: Obsidian
Free. Everything stays on your computer. No subscription, no mandatory cloud. It is what we use here.
09

What I built
during my internship

Four working AI tools. Each solves a real Verrier problem — designed, tested and deployed during the internship.

🧩
What is a “skill”?
A skill is a small program I wrote with AI to automate one specific task. You give it an input — a description, a file, a recording — and it returns a directly usable result. No software to buy, no subscription: it runs locally on Verrier’s computer.
Skill 01
🏷️
Product coding
A customer sends an email with 20 items to be coded by hand. The skill reads the descriptions, identifies each product family and returns the right code — automatically, in seconds.
83,9% accuracy validated against the in-house expert
→ Semantic embeddings (e5-small)
→ LLM rerank on ambiguous cases
Demo — a real email to paste into Claude Code
Email received — Sophie MARTIN, Tilleuls School
From: contact@college-exemple.fr
To: Ms D. — Verrier
Subject: Quote request + delivery follow-up
Dear Ms D.,

i hope you are well. While i'm requesting a quote, could you also let me know where we are with the delivery of the office chairs so we can schedule the “fitting” date.

Could you please prepare a quote for:

· 4 × 16-month academic diaries
· 1 × 30cm ruler
· 1 × black ink refill for Trodat Printy 4923 stamp
· 1 × projector remote
· 1 × box of 1000 copper 8mm staples
· 1 × basic 2-hole punch
· Rubber bands, wide + narrow
· BIC Velleda markers, blue / black / red (×12 each)
· Pack of 12 AA + pack of 12 AAA batteries
· 4 × essential-oil air fresheners
· 1 × wireless presentation remote
· 1 × pack of 100 punched pockets
· 1 × 10-drawer filing unit
· 1 × pack of 100 blue leather-grain covers
· 1 × wireless AZERTY keyboard

With thanks.

-- Regards, Sophie MARTIN · General Secretary · Tilleuls School · —
Bonus tool
🎯 Tender radar
Monitors public tenders in real time — never miss a relevant one again.
Open ↗
Bonus tool
✦ Verrier Copilot
An AI assistant built for Verrier — it knows the products, the processes, the customers.
Open ↗
Outlook · Horizon 2030

Five years from now.

Companies that will have put AI at the core of their operations won’t just be more efficient — they will be playing in a different league.

🧠
2026 — Now
The first tools are in place
Product coding, tenders, retention, transcription. Repetitive tasks are starting to be handed over to AI.
🔗
2027 — Knowledge base
The AI knows Verrier by heart
A second brain shared across the whole team. New joiners are up to speed in a day. No information gets lost.
⚙️
2028 — Automation
Agents that work without being asked
AI monitors tenders, follows up with at-risk customers, prepares quotes. People approve. The admin load disappears.
2030 — The real advantage
People do what AI
cannot do.
Customer relationships, commercial instinct, creativity, strategic judgement. The companies that win are not the ones with the most people — they are the ones that delegated everything else.
🏆 Lasting competitive advantage
⚡ 10× faster to respond
🎯 Team 100% on high-value work
The question is no longer “will AI change our trade?”
It is already happening.
The only real question is whether we are in it, or watching others move ahead.