27 Pages in 20 Minutes
The Problem
Writing Scope of Work documents for major exhibitions takes a full working day (8+ hours). The process involves gathering specifications, structuring sections, cross-referencing brand guidelines, and producing dozens of pages of technical content — all manually.
Before
- Full working day (8+ hours) per document
- Manual research of brand guidelines
- No structured review process
- Inconsistent formatting across sections
- Single-person bottleneck
After
- Complete 27-page SOW in 20 minutes
- AI retrieved brand guidelines autonomously
- Dual-agent review (vendor + security)
- Consistent professional formatting
- Identified gaps human missed
What Happened
Claude Code in plan mode asked diagnostic questions about the exhibition — year, square footage, specifications, brand requirements. It then autonomously retrieved the organization's standard format and generated 27 pages of structured SOW content. The key moment: two parallel reviewer agents were spawned — one assessing the document from a potential vendor's perspective, the other auditing security protocols. The security agent flagged missing VIP access management procedures that the human author hadn't considered.
Two Agents, One Midnight
The Problem
A manager assigned an urgent digital interaction performance analysis with a Monday deadline — a project that would normally take two to three weeks of research, data gathering, framework design, and presentation building. The professional needed a formal acknowledgment email, a performance analysis framework, presentation slides, and supporting data structures.
What Happened
In a midnight session, Claude Code spawned two parallel agents — one crafting the professional acknowledgment email with explicit deliverable confirmation and a proactive alignment meeting request, the other building the presentation structure with slide frameworks. Observers watching the live session took photos and screenshots to share, reacting in Arabic as slides generated in real time.
"Everything's ready — exactly what we needed."
From Voice Notes to Professional Reports
The Problem
An executive coach was spending 2–3 hours after every session manually writing coaching reports. She would take notes during sessions, then sit at her laptop typing up structured reports for each client — summarizing what was discussed, what goals were set, and what progress was made. This admin work was eating into her capacity to take on new clients.
Before
- Handwritten notes during sessions
- 2–3 hours writing each report manually
- Scheduling via WhatsApp back-and-forth
- No client history tracking
- Reports inconsistent in format
After
- Records session audio on her phone
- AI generates structured reports in minutes
- Clients self-book via integrated calendar
- Full client and session history dashboard
- Professional, consistent PDF reports
What I Built
A complete web application purpose-built for her coaching workflow. The app handles the full session lifecycle — from client booking to AI-generated report delivery.
- Voice-to-Report Pipeline — Upload a session recording (any format, up to 200MB). OpenAI Whisper transcribes it, then Claude AI generates a structured coaching report with goals, observations, and action items.
- Calendar Booking System — Integrated Cal.com scheduling. Clients book directly; webhooks auto-create session records. No more WhatsApp scheduling.
- Client Management Dashboard — Full client database with session history, report status tracking, and one-click report delivery.
- PDF Reports — Professional coaching reports that work on any device including iOS, with print-optimized CSS formatting.
- Security Hardening — HMAC webhook verification, rate limiting, timing-safe authentication, input sanitization, XSS prevention. Passed a 4-vector security audit.
- Large File Handling — Vercel Blob integration for audio files over 4.5MB, bypassing serverless platform limits without the client noticing.
"I just record the session on my phone, upload it, and the report writes itself. I used to dread the paperwork — now I actually have time for more clients."
Board Deck, Five Keystrokes
The Problem
A foundation director needed a board of directors presentation incorporating real organizational data — board members, recent exhibitions, institutional milestones. Gathering this information and structuring it into a professional deck typically takes hours of research and formatting.
What Happened
During a live demo, Claude Code's plan mode autonomously navigated to the organization's website, identified board members, pulled exhibition data, and built a targeted presentation — all from a single prompt. The AI demonstrated genuine research capability, not just text generation.
From Listing Site to Revenue Platform
The Problem
A villa and farm rental startup in the UAE had a live website built on Next.js, but was struggling with poor search visibility and had no way to handle payments, host payouts, or track revenue. The founder wanted to evolve from a simple listing directory into an Airbnb-style marketplace but didn't know where to start technically.
Phase 1: Technical SEO Audit
Ran a comprehensive 28-point technical audit using automated browser testing and API analysis. The site scored 9 out of 25 — revealing critical issues the founder had no idea existed.
- HTTPS not enforced — HTTP traffic served without redirect; Google penalizes this
- 4 duplicate domains active — http/https × www/non-www all returning content with no canonical tags
- No XML sitemap — Googlebot couldn't systematically discover listing pages
- 3.1s server response time — aggressive no-cache policy forcing every request back to origin server
- Zero structured data — missing JSON-LD schema that drives rich results in Google
- Missing security headers — no HSTS, no X-Frame-Options, weak CSP
Delivered a prioritized 5-page PDF report with exact fix instructions — the HTTPS fix alone takes 5 minutes via a single Cloudflare toggle.
Phase 2: Marketplace Architecture Blueprint
Designed the complete technical architecture for transforming the listing site into a revenue-generating marketplace platform.
- Revenue model — 15% platform commission, 85% to hosts. UAE 5% VAT applied to commission only. All amounts stored in fils (1/100 AED) to avoid floating-point errors.
- Database schema — 11 tables covering users, host profiles, properties, bookings, financial records, payouts, notifications, reviews, and audit log.
- Host dashboard — Earnings summaries, property-level transactions, payout history with line items, bank settings with IBAN validation.
- Admin dashboard — KPI cards, booking funnel, revenue analytics with VAT column, payout approval queue, host management.
- Automated payouts — Weekly cron job aggregates completed bookings per host. Idempotent by design — UNIQUE constraints prevent double-counting.
- UAE-specific compliance — TDRA data residency, UAE IBAN format validation (MOD-97 checksum), WhatsApp as primary notification channel.
"I had no idea HTTP wasn't redirecting to HTTPS. And the marketplace blueprint gave me a complete roadmap — I can hand this directly to a development team."
An AI-Powered Personal Operating System
Why This Matters
I don't just consult on AI — I live inside AI systems I built for myself. This case study is included because it demonstrates that what I teach is what I practice. Every recommendation I make to clients comes from systems I've tested on my own life first.
System 1: Automated Daily Intelligence
I wear a Limitless AI Pendant that captures conversations throughout my day. Every morning at 7 AM, an automated pipeline processes the previous day's data:
- Data capture — Limitless Pendant records 30–50 conversation logs per day (~1.6M characters)
- Processing pipeline — n8n automation fetches logs, condenses transcripts, sends to Claude Sonnet for analysis
- Output — Structured daily insights (key decisions, action items, topics, follow-ups) pushed to Notion and to my personal AI assistant's memory
- Result — I start every day with a briefing of what happened yesterday, what I committed to, and what needs attention
System 2: Notion Command Center
A PARA-method second brain with 7 interconnected databases:
- Tasks, Projects, Areas, Resources, Notes, Read Later, Daily Pages — all cross-linked with dual relations
- Task system — Status tracking, due dates, area/project linking, energy levels, recurring task support
- Automated population — Daily pages generated automatically, tasks synced from multiple sources
System 3: Personal AI Assistant ("Sam")
A custom AI assistant running 24/7 on a cloud server, communicating via Telegram:
- Always available — responds to voice and text messages at any hour
- Persistent memory — maintains context across conversations, remembers past decisions and preferences
- Integrated — reads email, accesses daily insights, checks calendar, manages follow-ups
- Cost-optimized — switched from premium to efficient AI model, reducing costs by ~$500/month while maintaining quality
Why I Do This
I'm an Avionics Engineer by training — I studied how to build systems where failure isn't an option. That mindset shaped every stage of my career: leading large-scale IT infrastructure projects at DP World (ports and logistics), Majid Al Futtaim Shopping Malls (retail technology), and the Department of Culture & Tourism Abu Dhabi (government digital transformation).
Across these roles, the pattern was always the same: smart people drowning in manual processes, while technology that could help them sat unused or was badly implemented. The gap was never the technology — it was the translation layer between what the tools could do and how people actually worked.
When AI tools became genuinely useful in 2024–2025, I saw the same gap opening again — but bigger. People hear about AI everywhere. They know it's important. But they don't know how to make it work for their specific job, their specific workflow.
So I started showing people around me — my wife, my cousins, my colleagues. One by one. Sitting with them, understanding what they do, and building or demonstrating solutions right there. The results were immediate and the reactions were the same every time: "Why didn't I know I could do this?"
That's what I do now. I bring an engineer's discipline to AI enablement — not selling hype, but building real systems that work on day one.
How I Actually Teach
These aren't principles from a training manual. They come from how I naturally work with people — refined across dozens of real coaching sessions. Here's what you can expect when we work together.
We Start with Your Problem, Not the Technology
You won't hear a pitch about AI. Instead, I'll ask: "What's taking you the most time right now?" We focus on your pain point first — the technology only appears when you see it solving your actual problem.
You'll See Results Before You Commit
Skeptical? Good. You don't need to believe me — just try one thing with me. When you see your own work done in seconds, the skepticism dissolves on its own. No convincing required.
We Start Small, Then Scale What Works
You won't be overwhelmed. We pick one simple, high-impact task first. Once it works and you trust it, you'll naturally start asking "what else can it do?" — and that's when we expand.
You Watch It Happen Live
No slides. No decks. I share my screen and build the solution in front of you, explaining every step. You see the full process — not a polished demo, but the real thing with your real data.
You'll Understand What to Use and Why
You don't need to know how AI models work. You need to know: "this tool handles your reports, that one manages your calendar, and here's which one fits your job." I make the choices clear and simple.
Everything I Recommend, I Use Myself
When I tell you something will save you two hours, it's because it saves me two hours — and I can show you exactly how on my own screen. No theory. Only tools I've built and tested on my own life.
You'll Get Honest Guidance, Not a Sales Pitch
If AI isn't the right solution for your problem, I'll tell you. If a spreadsheet is enough, I'll say so. My goal is that you walk away more productive — whether or not you need me again after that.
How I Work
-
You Show Me How You Work
We start by looking at your actual workflow — not what you think you do, but what you really do day-to-day. You'll be surprised how many 30-minute tasks are hiding in your routine that a machine could handle in 30 seconds.
-
Together We Find the Biggest Wins
Not everything in your work needs AI. I'll be straight with you about what's worth automating and what isn't. We look for the tasks that are frequent, time-consuming, and well-suited for AI — and ignore the rest.
-
You See It Working — On Your Screen, With Your Data
By the end of our first session, you'll have seen a real solution running on your actual work. Not a pitch deck, not a generic demo — your data, your workflow, your screen.
-
You Decide: I Build It, or I Teach You
Some people want me to build the full solution. Others want to learn how to do it themselves. Either way, the goal is the same: you walk away independently productive with AI.
What I Offer
AI Coaching Sessions
1-on-1 sessions where I learn your workflow and show you how AI fits into it — using your actual work, not generic demos.
Workflow Automation
Set up AI-powered pipelines, templates, and integrations that save you hours every week. Notion, email, documents, data processing.
AI Advisory Sessions
Hands-on 1-on-1 sessions where I diagnose your workflow and demonstrate AI solutions using your actual work. You'll see results before you commit to anything.
Let's talk about your workflow.
The first conversation is free. I'll tell you honestly
whether AI can help — and if it can, exactly how.