Le Pearl, run by an AI workforce.
Le Pearl Fashion is a real family-owned fashion retailer in the Gulf — 27 stores across the UAE, Oman & Saudi Arabia, since 2008. This is a live teaching demo of the 13 AI agents that can run it: shooting the catalogue, running the shop, app, social & WhatsApp, hiring staff, building the website by prompts, and talking to customers. Made to show students exactly how AI works.
What is Le Pearl?
Le Pearl Fashion is a real fashion retailer in the Gulf — a family business that runs 27 shops across the UAE, Oman and Saudi Arabia. It sells clothes for the whole family (infants, kids, men, ladies), plus footwear and accessories, and has dressed 50,000+ families since 2008.
How is AI used in this fashion store?
Running shops, a website, an app and social media usually needs a big team. Le Pearl instead uses 13 AI agents — software workers that each do one job: photograph clothes on AI models, write the product pages, run the store & app, reply on WhatsApp, post to social, screen job applicants, and even build the website from plain-English prompts. So a small team can run a big retailer. This page shows each agent working — live.
You don’t program it. You show it once.
Before the agents go to work, here’s the core idea every student should get: a modern AI agent learns a job by watching you do it once — then repeats it forever, on its own. No code, no prompts every time. This is the Stackbirds engine behind the whole demo.
Watch
You do the task once on your screen — the agent records every click, field and decision.
Learn
It self-trains on your exact process, tone, rules and edge cases — no prompt engineering from you.
Run
It repeats the job 24/7, accurately, and hands anything unusual back to a human. That’s an AI agent.
How Claude automates your work.
Every agent in this demo runs on Claude — Anthropic’s AI. It reads, decides and acts across your real tools: your store, your ERP, your inbox & calendar, WhatsApp, and the browser itself — the same way a person would, only tireless and instant.
13 agents. Five jobs to be done.
Every part of running Le Pearl maps to an agent with a name and one clear job. Tap See how it works for the plain-English steps, then Watch Claude do it — the exact prompt and Claude working in a live screen: filling the ERP, building a QuickBooks P&L from Excel, drafting Gmail replies, and more.
Try Scout yourself — real CVs in, ranked shortlist out.
This is the actual extraction tool behind the Scout agent above. Paste a candidate’s CV text, or drop PDF / DOCX / TXT files (or use the 22 samples) and watch it parse, score and rank every candidate against the job — everything runs privately in your browser.
Claude uses the browser like a person.
This is Claude in Chrome — the same side-panel you know. Hand it a supplier invoice and watch it open the ERP, move the cursor and fill every field itself. “Ask before acting” keeps a human in control.
It reads your email & calendar, and replies.
Watch Claude check the inbox, look at the calendar for a free slot, and write the replies — every step shown slowly, with a human able to approve before anything sends.
One flat photo → a model wearing it.
Drag the slider on each card: left is the plain flat-lay the supplier sends; right is the AI-generated on-model shot — same garment, brand-matched model, studio light. No photoshoot, no model casting.
Upload one photo → a published drop.
The same studio engine, running live in the app: upload a single fashion flat-lay and watch the on-model lookbook, the price-tagged listing and the published post across store, Instagram, WhatsApp & Telegram come to life.
Upload one photo. Get a lookbook.
A modest-fashion editorial reel — flat-lay to model shoot to published campaign — the same flow you just dragged in the studio above, in motion.
↳ Nano Banana MCP (Gemini) · generating frames… ✓
↳ matching audio · 12s · 1080×1920 · H.264 ✓
↳ rendering reel.mp4 … ✓
✓ reel.mp4 ready — playing on the phone above ▲
One prompt → a finished reel with audio.
The video layer that feeds every social post — generated in Claude Code via the Nano Banana MCP server (Gemini), then published with Blotato.
Install the Nano Banana MCP server & add your Gemini API key (one-time).
In Claude Code, describe the clip: subject, mood, aspect 9:16, and the audio.
Claude returns an image + matching audio — a ready-to-post short clip.
Drop it into Blotato and publish or schedule across every channel.
A store that runs and reports itself.
The product images above flow straight into a Shopify-style storefront — and the AI keeps the admin dashboard live: sales, orders, conversion, stock. Real Le Pearl products, real brands.
The Family Edit.
Infants to ladies, footwear & accessories — shot, priced & published by AI.
Bestseller
Low stock
Bestseller
Low stockThe same store, in your pocket.
An AI-curated feed, rich product pages, and a built-in stylist — every screen filled by the same agents that run the website.
AED 449AED 129AED 299AED 89You can build your own app — no coding required. Just prompt it.
↳ screens: Feed · Product (size picker) · AI Stylist chat · Cart & checkout
↳ wire to the store & payments API · push notifications · loyalty points
scaffolding React Native · components · tests … ✓
building TestFlight + Play Console builds … ✓
✓ Le Pearl app shipped to the App Store & Play Store — 0 lines hand-coded
The app was built the same way as the website — described in plain English to Claude Code. No code knowledge needed.
Every drop, broadcast in one tap.
In the Gulf, WhatsApp is the shopfront. The agents turn each collection into a WhatsApp campaign — a short Le Pearl marketing video, a branded template and a “Shop now” button.
A ready-to-send marketing template.
Pick a goal and the message rewrites itself for the drop — caption, media & CTA. The same short Le Pearl reel plays in-feed, in stories and as the WhatsApp broadcast preview.
“Email is where customers archive you. WhatsApp is where they reply.”
The full WhatsApp CRM & AI assistant — the chat that greets, answers, books and sends the pay-link. It plays automatically below.
How le-pearl.com was built by talking.
The whole site — this page included — was built by describing it in plain English to an AI (Claude Code), not by hand-coding. Watch the prompt become a live page.
Found by Google, and by AI.
The website & app aren’t just built — they’re made findable. One prompt adds classic SEO, answer-engine (AEO) and generative-engine (GEO) optimisation. Here is the prompt and the real code Claude writes.
↳ writing meta title/description · Open Graph · Twitter cards ✓
↳ JSON-LD: Organization · Store · Product · FAQ schema ✓
↳ AEO: Q&A blocks + concise answers for AI search ✓
↳ GEO: per-city pages (Muscat · Riyadh · Dubai) + geo tags ✓
✓ submitted sitemap · indexable by Google, Perplexity, ChatGPT & Gemini
<meta name="description" content="Shop Le Pearl — 27 GCC stores…">
<script type="application/ld+json">
{ "@type":"Store", "name":"Le Pearl Fashion",
"areaServed":["OM","SA","AE"],
"makesOffer":{"@type":"Product","name":"Aubergine Abaya"} }
</script>
<!-- AEO --> <section itemscope> Q: Do you deliver in Oman? A: Yes…
✓ Lighthouse SEO 100 · rich-results valid
No code knowledge needed — just prompts.
The exact “build a pro website with Claude Code” screen — the real setup & prompts behind le-pearl.com.
One agent runs the whole hire.
Screening is just one step. The full recruiting agent writes the job description, posts it to Indeed & LinkedIn, screens & ranks the candidates, sends the top picks their project details, and emails the owner a daily report — the demo below steps through every screen automatically.
Meet Layla — she talks, styles & books.
A live, auto-playing demo with real lip-sync and voice. Layla greets a customer, styles a look, takes the order and confirms on WhatsApp — then loops. Tap anywhere to enable her voice.
Enterprise infrastructure, multi-AI core.
A multi-layer stack: a backend with queues & webhooks, an AI engine layer combining several models, a media pipeline, and direct integrations with WhatsApp, Meta and the store.
Backend
- REST API architecture
- Relational store
- Queues & cache
- Webhooks & events
- Multi-tenant core
AI Engines
- GPT · text + vision
- Gemini · multimodal
- Computer vision
- Generative image AI
- Brand-locked gen
Media & CDN
- Cloud media manager
- Global CDN
- AI image processing
- Video processing
- Auto-compress · WebP
Channels
- WhatsApp Business API
- Meta Graph · IG / FB
- Telegram Bot API
- Ecommerce APIs
- Roadmap: Noon · Shopify
One product → every channel, tuned.
The Marketing agent takes one product and writes the right post for each platform — different tone, format and hashtags. Press the button and watch it publish to all six at once.
Daisy Tiered Girls Dress
One product photo + one idea — the agent does the rest.
The full 19-step social automation walkthrough.
This is the real playbook the Marketing agent follows — Claude Code + Blotato, screen by screen: generate posts, lock brand voice, make the visual, then post & schedule everywhere. It plays through below.