The Dictionary I Wish I Had When I Started
June 2025. 72 years old. 42 years in the travel business.
And I had no f***ing clue what a âpromptâ was.
I signed up for an AI course because I noticed my competitors were selling trips with chatbots while I was still using Excel 2003. On the first day, the teacher said: âWeâre going to optimize your prompts with few-shot learning techniques based on dataset context.â
I thought: âWrong class. This is for NASA engineers.â
Turns out it wasnât. Turns out these scary buzzwords are surprisingly simple when someone explains them without the jargon of a bearded hipster coding in a coffee shop.
This is what I wish Iâd known on day one.
The 5 Terms That Saved My Life
1. Prompting (or âhow to talk to the machine without pissing it offâ)
What it is: The way you talk to AI. Like ordering coffee from a waiter: if you say âcoffee,â you get a black espresso. If you say âlatte, hot, large cup, no sugar,â you get exactly what you want.
My real example (June 2025): â BAD: âGive me a ski tripâ â GOOD: âAct as a travel agent specializing in luxury skiing. Client: couple aged 40-50, intermediate level, budget âŹ8,000, prefer Swiss Alps, traveling in February, want 5* ski-in/ski-out hotel. Give me 3 options with itinerary, itemized prices, and aprĂšs-ski activities.â
What I learned: The more specific, the better. AI doesnât guess. Youâre the boss.
2. Training (or âteaching it your worldâ)
What it is: Feeding AI with YOUR specific information. Like onboarding a new salesperson: you hand them the catalogs, company policies, price lists. AI does the same thing, but in 30 seconds.
My real example (July 2025): I uploaded PDFs from 47 liveaboard boats (diving cruises) with itineraries, prices, and technical specs. I told ChatGPT: âYouâre a Red Sea liveaboard expert. Answer questions based ONLY on these documents.â
Result: An assistant that knew every cabin, every dive route, every high/low season price. Better than my own memory.
What I learned: AI without your data = generic. AI WITH your data = pure gold.
3. Memory (or âdonât forget who I amâ)
What it is: AIâs ability to remember previous conversations. Like when your doctor has your medical history and you donât have to remind them every single visit that youâre allergic to penicillin.
My real example (August 2025): I told ChatGPT about my ski holiday in Zermatt back in 1987 (yes, 39 years ago). Later, when I asked for help creating content about skiing in Switzerland, it generated text with references to âthe untouched magic of Zermatt youâve known for decades.â
I hadnât mentioned Zermatt again. It remembered.
What I learned: Memory turns AI from âtoolâ to âcoworker who actually knows you.â
4. Dataset (or âthe knowledge warehouseâ)
What it is: An organized collection of information that AI uses to learn or respond. Think of a massive library, but in digital form.
My real example (September 2025): I built a dataset from supplier emails (hotels, airlines, DMCs) with confirmations, prices, and cancellation policies. I trained AI on that dataset to automatically extract: hotel name, dates, room count, total price, conditions.
Before: 15 minutes reading each email. After: 10 seconds. AI reads, extracts, formats.
What I learned: Good dataset = effortless automation. Bad dataset = garbage processed at high speed.
5. Context (or âgiving it the right cluesâ)
What it is: The specific information you give AI in EACH conversation so it understands what youâre talking about. Itâs not what it remembers (thatâs memory) â itâs what you tell it right now.
My real example (October 2025): Without context: âWrite me an email to a clientâ With context: âWrite an email to Juan RodrĂguez, a client who booked a Maldives liveaboard in April, cancelled due to a family emergency, now wants to reschedule for October. Tone: empathetic, professional, offer 10% discount for the inconvenience.â
The first email: generic, flat, useless. The second email: perfect, human, straight to the point.
What I learned: Context = the difference between a clueless assistant and a brilliant one.
The Plot Twist
Today, 8 months after writing that âdictionary for the lostâ:
- I design custom GPTs that train my sales team
- I automate workflows with Make and n8n running in parallel
- I have 5 Claude Code agents working simultaneously on different tasks
- I process supplier emails in seconds with AI that extracts data and creates case files
Secret? None.
I just learned these 5 terms. And used them every single day.
For You, Just Getting Started
If youâre in June 2025 like I was (lost, overwhelmed, thinking âthis isnât for meâ), listen up:
Youâre not stupid. This is normal.
It took me 3 months to understand what a âdatasetâ was. Another 2 months to apply âcontextâ properly. And I didnât grasp the difference between âmemoryâ and âtrainingâ until September.
But once you master these 5 terms, everything else follows.
Because these 5 terms are the foundation of everything. The rest is just technical frosting.
PS for My Fellow 70-Somethings
Yes, it can be done. Iâm 72 and writing this from a Mac Mini that runs AI agents 24/7 while I sleep.
You donât need to be an engineer. You need curiosity and patience.
And a good dictionary.
Did this dictionary help? Share it with someone whoâs where you were 8 months ago. Sometimes all it takes is someone translating the jargon for you.
Giora Gilead Elenberg 72 years old | Entrepreneur since 1982 | AI learner since 2025 CICMA 2283 | Viajes Scibasku
What did you think?