CODED × Alshaya · Kuwait
CODED×ALSHAYA

The Tool Is Already
on Your Desk

Day 1 — Foundations, responsible use, and the framework that turns a vague request into a draft you can send.

DAY 1 Tuesday 1 September 2026 09:00 – 13:00 CODED Campus
Ali Taqi · Alshaya & Al-Injaz
press → to begin · ? for controls
Day 1

Yesterday's memo took you 45 minutes.

By the end of today, you'll draft one in under 10.

Before we begin
What is one task on your desk this week that you dread the most?
answers appear here
01
AI & LLM Foundations
🤖

What is AI, really?

Software that can understand, learn, and generate content that used to need a person to do it.
📖
Reads and understands text — a contract, a report, an email thread
✍️
Writes content — memos, notices, summaries, replies
💬
Answers questions — about anything you paste into it
🧩
Helps solve problems — options, trade-offs, next steps
🔋
Never gets tired and never forgets — at 8 AM or 8 PM, same quality
Think of it as a fast assistant, not a replacement for your judgment.
🧠

What is an LLM?

Large Language ModelIt read an enormous amount of text, learned the patterns in language, and now predicts what word should come next — one word at a time, until it has written a full answer.
It is not looking anything up. It is predicting the pattern — which is why it can be fluent and wrong at the same time.
The name

What makes an LLM large?

📚
Massive training data
Billions of pages of text — books, articles, code, conversation.
🔗
Billions of parameters
The internal settings it tuned while learning those patterns.
🏗️
Complex tasks
Enough scale to hold a long document in mind and reason across it.
🎯
One job
Predict the next word. Everything else is that, repeated.
Large is about scale, not intelligence. It is a very good pattern-predictor.
The landscape

Four names you'll hear

They all work the same way. What differs is who built them, and what each is strongest at.

ChatGPT
OpenAI
The best known. Strong all-rounder for writing and everyday questions, with a large add-on ecosystem.
Claude
Anthropic
Strongest on long documents, careful analysis, and natural professional writing. Built with a heavy focus on safety.
We use this
Gemini
Google
Built into Google Workspace. Convenient if your documents already live in Docs, Sheets, and Gmail.
DeepSeek
DeepSeek AI
An open, low-cost option from China. Capable, and popular where budget or self-hosting matters.
Learn the skill once and it moves with you. The prompt is the skill — not the logo.

Why should you care?

Not because it is new — because of what it gives back to your week.

⏱️
Save time
Automate repetitive writing. Summarise long documents. Get a first draft in minutes instead of an hour.
✅
Improve quality
Catch errors and inconsistencies, tighten unclear writing, and sound consistently professional.
💡
Think better
Brainstorm options, organise messy information, and see a problem from angles you had not considered.
📈
Work smarter
Read data faster, prepare for meetings quicker, and spend your hours on judgment instead of typing.
None of this replaces you. It removes the part of the job that was never the point.
Compare

What AI Does Well vs. Where It Fails

Weak at
Facts it cannot check, your company's private context, and final judgment calls
Strong at
Drafting, summarizing, brainstorming, and finding patterns in data
AI can state a wrong fact with full confidence. This is called a hallucination — always verify anything that matters.
🔒

What Can You Safely Type Into AI?

CautionDo not paste tenant records, financial data, or contract terms into a public AI tool. Use only information that is already public or approved for outside tools.
When you are not sure, ask your manager before you paste it.
Compare

Real Example — Writing a Memo

Before AI
45 minutes: blank page, first draft, two rewrites, final check
After AI
10 minutes: describe the memo, get a draft, edit, done
Same memo. One-quarter of the time.
Compare

Real Example — Reading a Long Report

Before AI
60 minutes: read the whole report and write notes
After AI
5 minutes: upload the report and ask for the key points
You still read the summary. AI reads the report.
Back to your list
Which one of these would save you the most time this month?
nothing captured yet — go back to slide 3
Two volunteers

Describe It, Draw It

One person sees the shape. One person draws it. The drawer faces the whiteboard, back to the screen — no peeking.

1
Pick two volunteers
One at the whiteboard with a marker. One facing the screen.
2
Describe it in words only
No gestures, no drawing in the air. The drawer may not ask questions — yet.
3
Two minutes on the clock
The shape is on the next slide. Press T for the timer, then compare the drawing to the original.
4
Run it again — questions allowed
Same shape, new pair. Watch how much faster it goes.
Describer only — drawer, eyes on the board
What just happened

Vague words in.
Vague drawing out.

✗
“Draw a cat on a wheel with a balloon.” — technically true, and the drawing looks nothing like it.
✓
“A square robot cat head with two triangle ears and an antenna, one X eye and one round eye, on a rectangular body, riding a single large wheel, holding a balloon in its raised right arm.”
The drawer is the AI. It can only draw what you actually described — it cannot see what is in your head.
02
Your First Prompts
Lab

Lab L1.1 — First Contact

8 minutes · in the Day 1 lab

1
Open the Day 1 lab — Task L1.1
It's linked from the Day 1 page. Your notes save in your browser.
2
Pick ONE starter prompt and run it in Claude
Eight starters in the lab, tap to copy. Any of them is fine — this is a confidence exercise.
3
Read the result closely
Would you actually use it? That is the bar. Note one thing that surprised you.
4
Finished early? Take the stretch
Steer the result without rewriting your prompt: “Make it shorter.” → “Make it warmer.”
Open lab task L1.1 ↗
Framework

What Makes a Good Prompt?

Specific — say exactly what you want
Clear purpose — say why you need it
Give details — background, numbers, names
Set the tone — formal, friendly, or technical
Compare

Weak Prompt vs. Strong Prompt

Weak
“Write something about the new parking policy.”
Strong
“Write a 100-word notice for tenants about the new parking policy starting Sunday. Use a friendly, clear tone.”
Break
Stretch and Reset
15:00
15 minutes — back on time, please
03
The CTFT Framework
Framework

CTFT: Four Parts of a Great Prompt

Context — the background
Task — the exact request
Format — how you want it structured
Tone — the style and voice

C = Context

Tell the AI your role, the situation, and who will read the result. AI does not know your building, your team, or your project unless you say so.
✗
“Write about the policy.”
Which policy? Whose? For whom?
✓
“I manage leasing for a retail property with 12 vacant units. Our tenant parking policy changes on Sunday.”
No context, no relevance. Start with who you are and what is going on.

T = Task

State the one action you want, written as a clear command. Ask for one thing, not five things at once.
✗
“Help me with this.”
Help how — write it, shorten it, check it?
✓
“Write a one-page leasing update for the regional director.”
One clear verb, one clear deliverable. Write, summarise, compare, list — pick one.

F = Format

Say how long the result should be, how it should be structured, and what type of document it is.
✗
“Write a document.”
How long? Bullets or prose? What sections?
✓
“One page. Opening summary, three bullet points, then a closing recommendation.”
If you do not choose the shape, the AI chooses it for you — and you rewrite it.

T = Tone

Say the style you want: formal, friendly, technical, or persuasive. The same facts land very differently.
✗
“Write an email.”
To a regulator, or to your team?
✓
“Professional and direct — this goes to the regional director, who is short on time.”
Tone is not decoration. It is how the reader decides whether to act.

Two Extra Tools: Role and Constraints

Role tells the AI who to act as — “act as a facilities manager.” Constraints set a limit — “do not mention the budget number.” Add these to CTFT when you need more control.
Optional extensions — use them only when plain CTFT is not specific enough.
Compare

One Task, Two Ways to Ask

Zero-shot
Ask directly, with no example: “Write a reply to a tenant complaint about noise.”
Few-shot
Give AI one sample reply first, then ask it to match that same style.
Use few-shot prompting when tone and format matter more than usual.
Compare

CTFT in Action

Poor prompt
“Write about the maintenance backlog.”
Excellent CTFT prompt
[CONTEXT] I manage facilities at a retail property with 40 open maintenance tickets. [TASK] Write a one-page update for the operations director. [FORMAT] Opening summary, three bullet points, closing recommendation. [TONE] Professional and direct.
Lab

Lab L1.2 — Build a CTFT Prompt

12 minutes · in the Day 1 lab

1
Choose your scenario in the lab
Scenario A (leasing) or Scenario B (maintenance) — both sample data.
2
Read the stand-in document, fill the four fields
Context, Task, Format, Tone — the lab assembles them into one prompt.
3
Copy the assembled prompt, run it in Claude
The lab has a Test in Claude button.
4
Compare the answer with the document
Did it miss anything? Did it invent anything? Flag both in the capture box.
Open lab task L1.2 ↗
04
Writing & Communication With AI
Framework

Three moves you'll use every day

Draft — turn a prompt into a first version
Rewrite — change the tone or the length of something that already exists
Summarize — pull the key points out of a thread or a document

1 — Draft from a prompt

You give Claude a CTFT prompt. It gives you a first version. That first version is never the final version — it is the thing you react to.
✗
“Write an email about the lift.”
Which lift, when, to whom, how long?
✓
[CONTEXT] I am the facilities lead at a retail property. The north lift is out Thursday, 9 AM to 12 PM. [TASK] Write a notice email to all tenants. [FORMAT] Subject line plus three short paragraphs, under 120 words. [TONE] Clear and apologetic, but brief.
A blank page costs you 45 minutes. A draft you can argue with costs 30 seconds.
Compare

2 — Rewrite: change the tone

Formal
“We wish to inform tenants that the north elevator will be out of service from 9 AM to 12 PM on Thursday.”
Conversational
“Quick note — the north elevator will be down for repairs Thursday morning, 9 to 12.”
Same facts, same length. Two different readers.
Rewrite

2 — Rewrite: change the length

You do not need a new prompt to fix a draft. Say what to change, in plain words, in the same conversation.
↓
“Make it half as long.”
↓
“Cut this to three bullet points.”
↑
“Expand the second paragraph — add the reason for the delay.”
→
“Say the same thing in one sentence.”
Shorter is almost always better. Ask for half, then read it again.

Refine, Don't Restart

When a draft is close but not right, tell Claude exactly what to change. “Make it shorter.” “Make it warmer.” Do not start a new prompt from zero.
This is called iterative prompting — small corrections, same conversation.

3 — Summarize the key points

Paste in a long email thread or a set of meeting notes. Then ask for the shape of summary you actually need — not just “summarize”.
✗
“Summarize this.”
You get a paragraph. You still do not know what you owe anyone.
✓
“Summarize this thread in five bullet points. List every decision made, every question still open, and who owns each action.”
Ask for the shape you need. Then check it against the original.
✉️

Email, end to end

Three things Claude handles well on email — and all three are just Draft, Rewrite and Summarize pointed at your inbox.
📥
Summarize a thread — “What was decided, what is still open, and what do I owe anyone?”
✍️
Draft a message — “Reply to the tenant below. Apologise for the delay, give Thursday as the new date, under 100 words.”
↩️
Match your own voice — “Reply in the same tone I used earlier in this thread.”
Right now this is copy and paste. The next slide removes the copy and paste.
One step further

Connect it straight to your inbox

MCP · Model Context ProtocolAn open standard that lets an AI assistant plug into the tools you already use — mail, calendar, files. Once it is connected, Claude reads the thread and drafts the reply in place. No copying, no pasting.
◆
Claude — connects to mail, calendar and files through MCP connectors
◆
ChatGPT — its own connectors for mail and documents
◆
Microsoft Copilot — built directly into Outlook, already inside Microsoft 365
This points an AI tool at real company mail. Nobody connects anything until IT and your manager have approved it.
Lab

Lab L1.3 — Draft a Professional Document

15 minutes · in the Day 1 lab

1
Pick a brief in the lab — or use your own real task
A staff memo, a tenant notice, or a status update. Tap to copy.
2
Write a full CTFT prompt for it
Do not skip Format or Tone.
3
Generate the draft, then edit by hand
Fix the tone, check every fact and number, cut anything repeated.
4
Read your final version out loud once
If you stumble, so will the reader.
Open lab task L1.3 ↗
Lab

Lab L1.4 — Rewrite It: Tone and Length

10 minutes · in the Day 1 lab

1
Copy the clunky paragraph from the lab into Claude
You already have a draft — now change it without starting over.
2
Run the six rewrite commands, in order
Same conversation. Do not start a new chat.
3
Notice what survives every version
That is the actual message.
4
Capture the version you would send
And note who you would send it to.
Open lab task L1.4 ↗
Break
Stretch and Reset
10:00
10 minutes — back on time, please
05
Reviewing AI Output & Using AI for Research

Before You Send Any AI Draft, Check:

Accuracy — are the facts correct?
Tone — does it match the reader?
Structure — is it easy to follow?
Numbers — did you verify every calculation by hand?

AI as a Research Assistant

Paste in a long document and ask for a summary, the key points, or the open questions it raises.
You still read the summary. AI reads the document.
Compare

Real Example — A 20-Page Report

Before AI
60 minutes: read the full report, take notes, write a summary
After AI
5 minutes: upload the report, ask for a summary, verify the key numbers
Lab

Lab L1.5 — Tame the Thread

10 minutes · in the Day 1 lab

1
Copy the five-email thread from the lab
A long thread you have been copied into — sample data.
2
Ask for the summary you actually need
Decisions made, questions still open, and who owns each action.
3
Check the summary against the thread
Did it miss anything? Did it invent an owner?
4
Draft your reply, then rewrite it once for tone
The reply goes to the regional director.
Open lab task L1.5 ↗
Lab

Lab L1.6 — Summarize and Verify

12 minutes · in the Day 1 lab

1
Download the quarterly report from the lab
Five pages of sample data — safe to upload. Attach it to Claude.
2
Ask for a one-page summary
Use a full CTFT prompt. Say what shape you need.
3
Check three numbers by hand
Against the report, not against the summary.
4
Find a planted mistake
There are five in there. One is enough.
Open lab task L1.6 ↗
Today you built:

Three things that are already yours.

A CTFT prompt for a real work task
A professional document, drafted and edited with Claude
A document summary, verified against the source
Tomorrow: turn today's skills into presentations and data insights.
End of Day 1

See you tomorrow.

Day 2 — Presentations, Data Analysis & Reusable Workflows

Wednesday 2 September 2026 · 09:00 – 13:00
‹ Day 1
Exercise2:00
1 / 0
→ next ← back F full ? help
screen blanked — press B to resume

Presenter & pointer controls

→SpacePg Dn
Next slide
←Pg Up
Back
L
Laser pointer (follows the mouse)
B.
Blank the screen
F
Fullscreen
S
Sound on / off (starts off)
T
Start the exercise timer (task length on a lab slide, otherwise 2:00)
?
This panel
Esc
Close / resume
Most USB presenter remotes send Page Up / Page Down — they already drive this deck.
☕ Start a break
Pick a length — the timer counts down and shows everyone the return time.
Break
Break
10:00
Back at —