← Swipe to navigate →

AI Masterclass for
Executive Leadership

From Zero to Vibe Coding — Build a Real AI Product in One Session

Cem Koşmaz & Erda Gerçek · March 2026
Corporate Teams · Full Day (09:00–17:00) · Hands-On Workshop
Agenda

Today's Journey

🧠

Part 1 — AI Foundations

What AI is, how it works, why it matters — for business leaders, not engineers

09:00 – 10:15💻 Cem + 🎓 Erda

Part 2 — Vibe Coding

"The hottest new programming language is English" — live demo: idea → working app

10:30 – 11:15💻 Cem
📊

Part 3 — AI Strategy Analyst

SWOT, competitive analysis, value chain — live strategic prompting for executives

11:15 – 12:00🎓 Erda
🚀

Part 4 — AI for Your Business

Apply strategic AI prompts to YOUR business problems with guided support

13:00 – 14:30🎓+💻 Both
🏗️

Part 5 — Build Your Own Tool

For the hands-on: build a simple AI-powered tool from scratch

14:30 – 15:30💻 Cem
🧭

Part 6 — Strategy & Q&A

On-prem vs cloud, corporate rollout, security & implementation roadmap

15:45 – 17:00🎓+💻 Both
01

AI Foundations

Understanding the technology before you invest in it

How LLMs Work

Next-Token Prediction

An LLM is a probability machine. Given a sequence of words, it predicts the most likely next word.

"The capital of France is"
Paris 92%
Lyon 3%
Nice 1%

🌡️ Temperature

Low = safe, predictable answers. High = creative, surprising. A "randomness dial" for the AI.

📏 Context Window

How much AI "remembers" per conversation. GPT-4: 128K tokens. Gemini 2.5: 1M tokens. Entire codebases fit.

Executive Takeaway

AI doesn't understand — it predicts. But prediction at scale, trained on the world's knowledge, produces something transformative.

The Future of AI

Intelligence as a Utility

💬
"We see a future where intelligence is a utility."

— SAM ALTMAN, CEO of OpenAI

⚡ Sold Like Electricity

Altman says AI will be sold like electricity and water — metered by usage. Compute will decide who gets access.

📄 A New Monthly Bill

In the near future, businesses will have a new utility bill to pay every month: artificial intelligence. Just like power and water.

Source: Business Insider · Sam Altman keynote

How AI Is Metered

AI is Metered in Tokens

Just like electricity is measured in kilowatts, AI is measured in tokens. This is the unit you will be billed on.

💡

Electricity

You consume kilowatt-hours.
Your meter counts usage.
You pay the bill monthly.

🤖

AI Intelligence

You consume tokens.
API meters every request.
You pay per token used.

Why We Start With Tokens

To understand AI costs, strategy, and real ROI — you need to understand what a token is. It's the atom of everything AI does.

Tokens

What is a Token?

AI doesn't read words — it reads tokens. A token is a chunk of text, roughly ¾ of a word.

Token ization splits text into pieces that AI can process
This sentence = 10 tokens · Each token costs money when you call an AI API

📏 Rule of Thumb

1 token ≈ 4 characters in English
1,000 tokens ≈ 750 words
A full page of text ≈ 500-800 tokens

💰 Why It Matters

You pay per token — both input (what you send) and output (what AI writes back). Understanding tokens = understanding costs.

How Context Works

Context Changes Everything

This Sunday I will go to ___
🏖️ the beach
⛪ church
⚽ the soccer game
🏠 my parents' house
🛒 the store
I play soccer every weekend. This Sunday I will go to ___
⚽ the soccer game
⚽ practice
🏖️ the beach
⛪ church
I play soccer every weekend. My team has a championship final this Sunday. This Sunday I will go to ___
⚽ the soccer game
🏆 the championship
🏖️ the beach
This is What a System Prompt Does

A system prompt gives AI context — exactly like adding "I play soccer" above. The more relevant context you provide, the more accurate and focused the AI's output becomes. This is why prompt engineering matters.

🔗 Transformer Explainer — See Probabilities Live
Token Pricing

What Does AI Actually Cost?

Per 1 million tokens · Prices as of March 2026

Provider Model Input / 1M Output / 1M Type
OpenAI GPT-5.4 Flagship $2.50 $15.00 🔒 Closed
Anthropic Claude Opus 4.6 Flagship $5.00 $25.00 🔒 Closed
Anthropic Claude Sonnet 4.6 Value $3.00 $15.00 🔒 Closed
Google Gemini 2.5 Pro Value $1.25 $10.00 🔒 Closed
DeepSeek DeepSeek V3.2 Open $0.28 $0.42 🔓 Open
DeepSeek DeepSeek R1 Reasoner $0.55 $2.19 🔓 Open
Alibaba Qwen 3.5 Plus Open $0.26 $1.56 🔓 Open
Key Insight

Open-source models are 10-60x cheaper than closed-source — and you can run them on your own hardware. The gap in quality is closing fast.

Real Usage Data

What We Actually Spend

Anthropic API · October 2025 · One month of production AI

855M
Input Tokens
3.3M
Output Tokens
$673
Total Monthly Cost
~$22
Average Daily Cost

📊 What This Runs

14 production AI applications including chatbots, code agents, content pipelines, and data analysis — all powered by Claude Sonnet 4.5

💡 The Comparison

$673/month for AI vs. $5,000+/month for one developer. AI handles the equivalent work of multiple specialists at a fraction of the cost.

Executive Takeaway

This is real production cost data. AI is not expensive — not using AI is expensive.

AI Coding Usage

Real AI Coding Usage — One Day

Cursor IDE · GPT-5.4 · Mar 14, 2026 · Every request = AI writing production code

Time Model Tokens API Cost Status
02:49 PMgpt-5.4-high393.4K$2.37✓ Included
02:29 PMgpt-5.4-high2.3M$9.70✓ Included
02:23 PMgpt-5.4-high1.5M$4.93✓ Included
02:19 PMgpt-5.4-medium1.1M$1.51✓ Included
01:48 PMgpt-5.4-medium8.1M$20.69✓ Included
01:46 PMgpt-5.4-medium573.7K$0.43✓ Included
01:44 PMgpt-5.4-medium272.4K$0.10✓ Included
01:42 PMgpt-5.4-medium1.4M$0.55✓ Included
01:39 PMgpt-5.4-medium2.1M$0.85✓ Included
01:37 PMgpt-5.4-medium756.1K$0.29✓ Included
01:36 PMgpt-5.4-medium486.8K$0.19✓ Included
01:29 PMgpt-5.4-medium712.6K$0.31✓ Included
01:27 PMgpt-5.4-medium1.4M$0.53✓ Included
01:05 PMgpt-5.4-medium3.5M$1.92✓ Included
What You're Looking At

One agentic coding session used 8.1M tokens ($20.69)

Intelligence Benchmark

AI Intelligence Leaderboard

Artificial Analysis Intelligence Index v4.0 · 10 evaluations (GDPval-AA, τ²-Bench, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt) · March 2026

🔵 Gemini 3.1 Pro Preview
57
⬛ GPT-5.4 (high)
57
⬛ GPT-5.3 (video/high)
54
🟤 Claude Opus 4.6 (max)
53
🟤 Claude Sonnet 4.6 (max)
52
⬛ GPT-5.2 (high)
51
🔷 GLM-5
50
🟣 Grok 4.20 Beta D309
48
⬛ Kimi K2.5
47
🔵 Gemini 3 Flash
46
🟠 Qwen3.5 397B A17B
45
🔴 MiniMax-M2.5
42
🔵 DeepSeek V2.2
42
🟠 MiMo-V2-Flash (Feb 2026)
41
🟣 Grok 4.1 Fast
39
🟤 Claude 4.5 Haiku
37
🟢 NVIDIA Nemotron 3 Super
36
🟡 Nova 2.0 Pro Preview
36
🔵 Gemini 3.1 Flash Preview
34
⬛ gptcass-12DB (high)
33

💡 Reasoning models indicated by lightbulb · Source: artificialanalysis.ai

Coding Benchmark

AI Coding Leaderboard

Artificial Analysis Coding Index · Terminal-Bench Hard + SciCode · March 2026

⬛ GPT-5.4 (high)
57
🔵 Gemini 3.1 Pro Preview
56
⬛ GPT-5.3 (high)
53
🟤 Claude Sonnet 4.6 (max)
51
⬛ GPT-5.2 (high)
49
🟤 Claude Opus 4.6 (max)
48
🔷 GLM-5
44
🔵 Gemini 3 Flash
43
🟣 Grok 4.20 Beta D309
42
🟠 Qwen3.5 397B A17B
41
🔵 Kimi K2.5
40
🔴 MiniMax-M2.5
37
🔵 DeepSeek V3.2
37
🟠 MiMc-V2-Flash (Feb 2026)
33
🟤 Claude 4.5 Haiku
33
🟢 NVIDIA Nemotron 3 Super
31
🟣 Grok 4.1
31
🟡 Nova 2.0 Pro Preview
30

Source: artificialanalysis.ai · Coding Index = Terminal-Bench Hard + SciCode weighted average · 💡 = reasoning model

The Speed of AI

How Fast AI is Improving

Intelligence Index scores · Top model per company · 2024 → 2025 → 2026

🟢 OpenAI

182024
352025
572026

🟤 Anthropic

162024
332025
532026

🔵 Google

152024
312025
572026
⚖️ US Bar Exam (Uniform Bar Examination)
2023: GPT-3.5 scored bottom 10% of test-takers · 2024: GPT-4 scored 90th percentile (297/400) · 2026: GPT-5.4 passes with near-perfect score — outperforming 99% of human lawyers
February 2026

Claude Cowork — AI as a Colleague

Released Feb 10, 2026 · Claude reads, edits, and creates files on your computer autonomously

📁 What is Cowork?

Claude's desktop app gets direct access to designated folders on your machine. It reads, edits, creates files — PDFs, Excel, Word, Markdown — without copy-paste. Like giving AI a desk in your office.

📝 CLAUDE.md Files

The key innovation: place a CLAUDE.md file in any project folder. Claude reads it as standing instructions — your company rules, tone, context — automatically applied to every task in that folder.

🔄 Cross-App Continuity

Move between Excel → PowerPoint → Word with full context retained. Analyze earnings, update financial models, build CIM slides — all in one seamless workflow without losing context.

💼 Enterprise Workflows

Entire deal workflows run autonomously: Claude reads NDAs, extracts deal terms, populates CRM fields, drafts memos — completing tasks that took hours in minutes.

The Shift

AI moves from chat tool to autonomous colleague — working inside your actual files, not just answering questions.

Finance AI Revolution

Claude's Finance Plugins

Released Feb 2026 · 5 professional plugins · Real institutional data integrations

🏦 Investment Banking

Reviews transaction docs, builds comps, prepares pitch books, drafts CIMs, teasers and process letters. Automates deal prep.

📊 Equity Research

Parses earning transcripts, updates financial models, drafts research notes. Reduces note prep from 6 hours → 45 minutes.

🔍 Private Equity

Automates diligence: reviews large doc sets, extracts standardized financial data, models scenarios across portfolio companies.

💰 Wealth Management

Analyzes portfolios, identifies drift and tax exposure, generates rebalancing recommendations with full client context.

📈 Financial Analysis

1M-token context window reads entire corporate document libraries in one session — 10-Ks, earnings, filings, all at once.

🔗 Live Data Sources

Connected to: LSEG · FactSet · MSCI · S&P Capital IQ · Bloomberg · Daloopa · Morningstar · PitchBook

For Your Business

The same plugin framework applies to any industry — legal, healthcare, logistics, HR. Claude in February 2026 became a domain specialist, not just a chatbot.

Model Sizes

Bigger ≠ Always Better

Size Parameters Analogy Best For
Small 1 – 8 Billion 🧑‍💼 Smart intern Fast, cheap, specific tasks
Medium 30 – 70 Billion 📊 Senior analyst Solid reasoning, most tasks
Frontier Undisclosed 🏛️ Consulting team Complex analysis, top quality
Key Insight

A well-tuned 8B model can outperform a 70B model on specific tasks. The right model depends on the use case and budget.

Training vs. Inference

Building the Factory vs. Driving the Car

🏭

Training

Cost: $2M – $200M+
Time: Weeks to months
Hardware: Thousands of GPUs
Who: OpenAI, Google, Meta

🚗

Inference

Cost: $0.001 – $0.10 per query
Time: Milliseconds
Hardware: 1 GPU or cloud API
Who: You (right now)

Executive Takeaway

You don't need to train models. You use them. Training is building the factory — inference is driving the car.

Closed vs. Open Source

Two Worlds of AI

🔒 Closed Source

GPT-4, Claude, Gemini

✅ Best quality
✅ Easy to start
⚠️ Data goes to their cloud
⚠️ Pay per token

🔓 Open Source

Llama 3, Mistral, Qwen

✅ Free to use
✅ Data stays on your hardware
✅ Full customization
⚠️ Need your own GPU

Real Example

Our SafetyEye runs local open-source models on our GPU server — factory data never leaves the premises.

RAG

Your Company's AI Brain

AI knows the world, but not YOUR company. RAG bridges the gap.

User Question
Search Your Docs
Feed to LLM
Answer with YOUR Data
"What were our Q4 margins?" → RAG searches your financial docs → AI answers with your real data. No expensive training required.
Multimodal AI

AI That Sees, Speaks, and Creates

👁️

Vision

Analyze images, documents, charts, factory floors in real-time

🎨

Generation

Create images, video, presentations from text descriptions

🗣️

Voice

Text-to-speech, meeting transcription, voice assistants

We Built These

Image Studio (GPU image gen) · Video Pipeline (auto video) · VibeVoice TTS (podcast audio) · SafetyEye (vision AI)

03

AI as Your Strategy Analyst

Your AI-powered boardroom advisor — live strategic prompting

Strategic AI Prompt

The Competitive Landscape Summary

You are a strategy analyst who has spent three weeks building a competitive intelligence report. Now compress everything into one executive-ready slide with no filler. I need: - Top competitors: The 8 most relevant competitors, ranked by how much they matter - Comparison dimensions: 6-8 real competitive factors (pricing model, target customer, distribution, type of advantage) - Competitor grid: Rate each competitor on each dimension with a short note - Open space: Positions that nobody currently owns — potential ways to stand out - Who is moving: Which competitors are gaining ground, stuck, or losing steam? - Biggest threat: The one player most likely to hurt my position in 18 months - Three-sentence summary: The competitive landscape in three sentences a CEO would read twice My company: [DESCRIBE YOUR OFFERING AND MARKET POSITION] Key competitors: [LIST WHO YOU CONSIDER COMPETITION]
🏢 Borusan Otomotiv Example: "Premium automotive distributor in Turkey — exclusive BMW, MINI, Jaguar, Land Rover. Competitors: Doğuş Otomotiv (VW/Audi/Porsche), Çelik Motor (Mercedes)..."
Strategic AI Prompt

The SWOT That Actually Works

You are a strategy consultant who refuses to produce the vague, generic SWOT. Give me a SWOT that is specific, backed by evidence, and directly useful. I need: - Strengths: 4-5 real advantages with proof — not generic claims, but specific things competitors cannot easily copy - Weaknesses: 4-5 honest internal problems, with a note on how hard each is to fix - Opportunities: 4-5 specific outside conditions creating real advantage windows, with timing - Threats: 4-5 outside forces that could hurt performance, with probability and impact rating - Interaction analysis: 4 most important S+O, S+T, W+O, W+T combinations and what move each suggests - "So what?" test: For every item in all four boxes, confirm it leads to a clear action. If not, remove it My business: [COMPANY, PRODUCT, OR INITIATIVE] Context: [WHAT DECISION THIS SWOT IS HELPING WITH]
⚡ Borusan EnBW Enerji Example: "855 MW installed renewable capacity (100% wind). Evaluating expansion into solar + EV charging infrastructure across Turkey..."
Strategic AI Prompt

The "So What?" Stress Test

You are a senior strategy partner reviewing an analyst's work before it goes to the client. Your only job is to keep asking "So what?" until every recommendation is solid. I need: - Recommendation check: Take each recommendation and ask "So what?" three times to push to concrete, actionable insight - Vague language detector: Flag every weak phrase like "improve efficiency," "drive growth," "optimize performance" — force into specific, measurable language - Logic chain check: Data → Insight → Implication → Recommendation → Action. Flag any broken links - Pushback rehearsal: What would a skeptical CFO or board member say? How should I respond? - Confidence level: Which recommendations have strong evidence vs. educated guesses? - Slide title test: If each recommendation were a slide headline, would it be specific enough to act on? My recommendations: [PASTE YOUR CURRENT RECOMMENDATIONS OR ANALYSIS]
💡 Universal: Paste any department's strategic recommendations — works for every business unit at Borusan. Try it with your own team's latest strategy deck.
Strategic AI Prompt

The Value Chain Breakdown

You are a strategy analyst mapping the full value chain to find where a business creates value, captures value, and loses value. I need: - Primary activities: Inbound logistics, operations, outbound, marketing/sales, service — assess each - Where we create value: Which activities genuinely create better value than competitors? - Where we lose value: Where does margin leak to suppliers, middlemen, or wasteful processes? - Competitor comparison: How does my chain differ from top competitor's? - Make vs. buy: Which activities could be outsourced without losing strategic control? - Biggest lever: The single activity that, if improved by 20%, would have the largest effect My business: [DESCRIBE YOUR COMPANY AND INDUSTRY] Biggest cost concern: [WHERE YOU FEEL THE MOST PRESSURE]
🏭 Borusan Pipe Example: "Leading steel pipe manufacturer in Turkey/Europe. Raw steel → pipe production → coating → logistics → B2B sales. Biggest pressure: energy costs and raw material price volatility."
Strategic AI Prompt

The Scenario Planning Matrix

You are a strategy partner running a scenario planning session for a leadership team facing serious uncertainty. I need: - Two key unknowns: The 2 most uncertain + impactful factors → these become axes of a 2x2 matrix - Four scenarios: Internally consistent stories (5-7 sentences each) for 3-5 years out - What each means for us: 3 specific effects on strategy, product, pricing, or operations per scenario - Safe bets vs. gambles: Moves that work across multiple scenarios vs. moves that only pay off in one - Early warning signs: 3 signals per scenario we can watch for TODAY - Decision triggers: Specific points where we switch from one approach to another My context: [DESCRIBE YOUR INDUSTRY AND SITUATION] Biggest uncertainty: [BIGGEST UNKNOWN YOU FACE]
🚜 Borusan Cat Example: "Caterpillar equipment distributor for construction/mining. Uncertainties: (1) speed of EV transition in heavy equipment, (2) Turkish construction sector growth trajectory..."
Strategic AI Prompt

The Operating Model Designer

You are an organization design consultant building the operating model that turns strategy into day-to-day execution. I need: - Strategy to model: How my stated strategy should shape every piece of the operating model - Structure options: Evaluate functional, divisional, matrix, and network structures — pros/cons for my case - Decision rights (RACI): For the 10 most important recurring decisions — who decides, who's consulted, who's informed - Key processes: 5 most critical processes for the strategy to work, and current gaps - Critical roles: 3-5 jobs that matter most for competitive advantage — what "great" looks like in each - Rollout order: What order to make changes in, to keep disruption low and get early wins My business: [COMPANY, STRATEGY, AND CURRENT SETUP] Biggest challenge: [WHERE STRATEGY IS NOT TURNING INTO RESULTS]
🏢 Borusan Holding Example: "6 business units (manufacturing, automotive, logistics, energy, machinery, ventures). Challenge: ensuring AI transformation happens consistently across all units, not just in tech-forward divisions."
Strategic AI Prompt — Master

The Full Strategy Synthesis

You are a senior strategy partner presenting final recommendations to a CEO and board after a six-week engagement. Pull everything together into one clear, decisive strategy document. I need: - Where we stand: Market position, competitive situation, financial health, team capability — precise and direct - Top 5 findings: Five most important things the analysis uncovered — sharp, evidence-backed insights - Pyramid narrative: Main message, 3 supporting arguments, 9 data points — clean pyramid structure - Three paths forward: Three strategic options with expected value, risk, capabilities needed, timeline - My recommendation: One clear choice with reasoning a skeptical board member needs - 30-60-90 day plan: Specific actions in first 30 days, progress check at 60, key turning point at 90 - One-page board summary: Headline, 3 findings, recommendation, and next steps on one page My business: [FULL CONTEXT: COMPANY, MARKET, STAGE, TEAM, REVENUE, BIGGEST CHALLENGE] Decision to make: [THE STRATEGIC CHOICE THAT NEEDS TO BE RESOLVED]
🎯 The prompt that replaces a 6-week consulting engagement. Try it with your business unit's biggest strategic question. The AI won't give you the answer — but it will give you the right structure to find it.
Your Toolkit

Strategy Prompt Toolkit

7 executive-grade AI prompts — copy to ChatGPT, Claude, or Gemini. Click any prompt slide to see it in full.

02

Vibe Coding

The hottest new programming language is English

The New Paradigm

What is Vibe Coding?

"The hottest new programming language is English." — Andrej Karpathy, ex-Tesla AI Lead, OpenAI Co-founder
Describe in English
AI Plans
AI Codes
AI Debugs
Working Product ✅

🤖 Antigravity (Gemini)

Google's agentic AI. Reads entire codebases, runs commands, builds full projects autonomously.

💻 Claude Code

Anthropic's terminal agent. Plans, codes, tests, and deploys — thinks before it acts.

Live Demo

Watch a complete app get built from a single English description

[SWITCH TO LIVE CODING ENVIRONMENT]

03

Instructor Live Build

Watch a complete AI application built from a single idea → 90 minutes

Hands-On Build

What We're Building

Define Problem
10 min
AI Plans
5 min
Build Backend
20 min
Build Frontend
20 min
Demo & Modify
35 min
Project Ideas

If You Haven't Decided Yet...

📈

Finance Research Assistant

Enter a company → AI researches financials, news, trends → generates executive briefing with charts

🏢

Competitive Intel Dashboard

Input industry + competitors → AI analyzes positioning, pricing → strategic comparison report

📋

Meeting Action Tracker

Paste notes → AI extracts decisions, action items, deadlines → auto-generates follow-up emails

05

Your Turn to Build

Think of a real problem from YOUR department. Now build it.

[PARTICIPANTS BUILD THEIR OWN PROJECTS — 60 MINUTES]

05

Strategy & Next Steps

What does this mean for YOUR business?

AI Playbook

Your Role. Your AI Opportunity.

Role Top Opportunity 30-Day Quick Win
CEO AI-first company strategy AI assistant for decision support
CMO Personalization & content at scale AI social media content pipeline
CFO Predictive financial modeling AI-powered cash flow forecasting
COO Process automation RPA + AI for invoice processing
CTO Vibe coding adoption Pilot on one internal tool
Our Portfolio

We Don't Just Teach AI — We Build It

🤖

Comturk AI

87% auto-resolution chatbot. WhatsApp, Telegram, Web.

👁️

SafetyEye

Real-time factory safety. YOLO + SAM3 on GPU.

📈

AI Forecaster

XGBoost + ARIMA. TCMB data. Automotive sales.

🏗️

SiteCraft AI

AI website builder. Agentic WordPress editing.

📊

SurveyAI

Conversational AI surveys. Google Forms migration.

🎬

Video Pipeline

Auto video generation. FFmpeg + AI content.

14 production AI projects · 1 GPU server · All built with vibe coding

Thank You.

You just built your own AI-powered product. Now go change your business.

Cem Koşmaz & Erda Gerçek
aipresentation.comturk.com
info@comturk.com
+90 212 996 7157 · +1 646 661 2912