Product Blueprint — April 2026

The Product They
Cannot Say No To

Every country in Singapore and the Gulf has spent $100M+ on cloud infrastructure that isn't producing intelligence. They all watched ChatGPT launch and asked the same question. You're the first person showing up with the answer.

$20M+
What Palantir AIP costs
$300K
Your pilot entry price
6 wks
To working demo with Claude
0 new
Cloud infra they need to buy
Sovereign AI Analyst — Live
What are the top 3 threat patterns on the northern border this month?
Based on 847 incident reports (Mar–Apr 2026), three patterns are elevated: (1) vehicle crossings between waypoints 14–17 increased 34% week-on-week. (2) Sensor blackouts at sector 9 correlate with subsequent incursions in 6 of 8 cases. (3) Activity peaks between 02:00–04:00 local, consistent with prior pattern from Q4 2025.
border_incidents.db sector9_sensors.csv patrol_logs_Q1.pdf +4 sources
The Problem Every Country Shares

Different flags, same crisis. Six countries, one unsolved problem worth hundreds of millions.

🗄️
Data Everywhere, Intelligence Nowhere
Border sensors, patrol logs, financial transactions, satellite feeds, customs databases — all in separate systems, none talking to each other. Analysts copy-paste between spreadsheets to write reports.
"We have petabytes of data and we don't know what's in it."
💸
Cloud Bought, Value Never Delivered
Kuwait signed with Microsoft. Qatar has Azure AND Google data centres. Saudi Arabia has Oracle. Singapore onboarded Oracle Cloud for DSTA. They paid for the pipes. No one built the intelligence layer on top.
"We have Azure. Our data still lives in Excel."
🤖
Ministers Saw ChatGPT. They Want That.
Every defence minister and intelligence chief has used ChatGPT. They all asked the same question to their IT teams: "Why can't we have this but for our own data?" Their IT teams have no answer. You do.
"Can we not just do what OpenAI did, but with our files?"
🔒
US Platforms Are a Sovereignty Problem
Palantir is US-origin. Gotham is ITAR-adjacent. Every country here has a reason not to send classified data to a US company server. Singapore is Five Eyes non-member. Gulf states have complex US political dynamics.
"We cannot put our military data in an American cloud."
⏳
Analyst Time Wasted on Manual Work
Senior intelligence officers spend 60–70% of their time reading, summarising, and copy-pasting between reports. The actual analytical thinking — pattern recognition, threat assessment — gets maybe 2 hours a day.
"My best analyst writes 4-page reports that no one reads."
🎯
Palantir's Price Is Unreachable
Palantir's minimum serious engagement starts at $10M–$20M. 18–24 month deployment. Requires their consultants embedded in your org. Bahrain, Kuwait, Qatar cannot justify this. Even Saudi wants an alternative for non-US-aligned work.
"Palantir wanted $35M and two years. We need something in 90 days."
Three Modules. One Product.

Each module solves a specific pain. All three run on their existing cloud. All three built with Claude API.

01
Ask Your Data
Natural language query over any government or defence dataset — the demo that closes the deal
2 weeksto live demo
What It Does
  • Connect their existing databases, Excel files, PDFs, API feeds
  • Any analyst types a question in English or Arabic
  • Claude reads relevant data, returns a sourced answer in seconds
  • Every answer cites which files/tables it used
  • Deployed entirely on their own Azure/Oracle — data never leaves
  • Role-based access: generals see military data, ministers see budget data
How You Build It
Claude API (claude-opus-4) RAG pipeline Vector database (pgvector) Next.js frontend PostgreSQL connectors PDF ingestion (pypdf) Excel parser REST API polling Azure / Oracle Cloud deploy Arabic NLP preprocessing
10 sec
vs 3 days for manual answer
100%
Sovereign — their cloud, their data
∞
Data sources connectable
0
New infrastructure needed
02
Automated Intelligence Brief
Every morning at 6am, every commander and minister gets a tailored briefing generated from overnight data — automatically
3 weeksto deploy
What It Does
  • Nightly job reads all new incident reports, sensor logs, news feeds
  • Claude extracts entities, patterns, anomalies, changes from baseline
  • Generates a structured 1-page brief per role: Commander / Minister / Analyst
  • Flags items requiring immediate attention vs. routine updates
  • Delivered by email, secure portal, or pushed to mobile app
  • Tracks what was briefed previously — no repetition, only deltas
How You Build It
Claude API (claude-sonnet) Cron job scheduler Template engine (structured output) Data diff pipeline Secure email delivery Role-based filtering PDF brief generator Baseline comparison logic
3–5
Analysts replaced per deployment
6am
Brief ready before first meeting
Zero
Overnight analyst shifts needed
100%
Coverage — no data missed
03
Document Intelligence
Upload any document — foreign language, captured material, intercepted comm, contract — and get an instant structured extraction
2 weeksto deploy
What It Does
  • Upload PDF, image, scanned doc, audio transcript, email chain
  • Claude extracts: people, places, dates, organisations, financial amounts
  • Auto-translates from Arabic, Farsi, Russian, Chinese, French
  • Generates: executive summary, key findings, recommended follow-up actions
  • Tags document into existing knowledge graph (links to related entities)
  • Flags if entities appear in watch lists or prior intelligence files
How You Build It
Claude API (claude-opus-4 vision) Structured output (JSON schema) Entity extraction pipeline OCR preprocessing Translation layer Knowledge graph (Neo4j) Watch list matching Secure upload interface
45 sec
vs 4 hrs for manual processing
8+
Languages supported out of box
Auto
Links to related intelligence files
100%
Audit trail — every extraction logged
Why They Cannot Say No

Every objection has a structural answer baked into the product design.

Their Objection Your Answer Why It's Bulletproof
"We don't trust a US platform with our data" Runs on their cloud. Azure Kuwait, Oracle Singapore, GCP Qatar. Data never leaves their sovereign infrastructure. You're a wrapper, not a data processor. Claude API calls are stateless — no training on their data. Contractually provable.
"Palantir was too expensive and took too long" $300K pilot, 90 days to live. If it doesn't work, they haven't spent their annual software budget. You de-risk the decision to near-zero. Pilot pricing puts the decision below most procurement thresholds — no committee approval needed in several countries.
"We have no engineers to maintain this" It runs on their existing cloud team's stack. Azure DevOps, Oracle Cloud ops — they already have staff who know how to run it. No new hires. You provide 12-month support SLA. Their team trains on standard cloud tooling they already use.
"We already bought Microsoft / Oracle / G42" This is the missing intelligence layer on top of what they already bought. It makes their $100M cloud investment actually produce value. It's complementary, not competitive. Microsoft, Oracle, G42 all provide infrastructure — none provide this intelligence layer. No territory conflict.
"What if the AI makes a mistake?" Every answer cites its source. Analysts verify before acting — same as today. You're removing the retrieval work, not the human decision. Claude doesn't act, it informs. The "AI makes mistakes" objection assumes AI replaces humans. This product assists humans. Every output is auditable.
"We need to see it work first" You build the demo in 2 weeks with synthetic but realistic data matching their specific country. Oil pipeline data for Kuwait. Maritime vessel logs for Bahrain. They walk out of the demo having used it. Demos built on their domain-specific data have a 10x close rate vs. generic demos. Claude builds the demo in days.
The Sales Motion

Four steps from cold approach to $2M–$8M annual contract.

1
Week 1–2
Build the Country-Specific Demo
Use Claude to build a working demo with synthetic but realistic data for their specific domain. For Kuwait: KNPC oil pipeline maintenance logs + border incident reports. For Singapore: SAF logistics + DIS threat reports. For Bahrain: port vessel data + patrol logs.

The demo must answer 3 questions they would actually ask in real life. Walk into the meeting and let them type their own question. They close themselves.
Cost to build: ~40 hours of your time with Claude
2
Week 3–4
Sign the 90-Day Pilot ($300K–$500K)
Pilot contract covers: connecting 2–3 of their real data sources, deploying on their cloud, training 5–10 analysts, and delivering 90 days of working usage. Fixed price. No overruns.

Keep the pilot scope tight: one ministry, one domain. Win there, expand later. Pilot priced below most government procurement thresholds — a director can sign it without committee.
Revenue: $300K–$500K
3
Month 4
Pilot Proves Value → Full Platform Contract
After 90 days, you have user adoption data, time-saved metrics, and internal advocates. Present the numbers: "Your analysts answered 340 questions in 90 days. Without this, each took 3 days. You got back 1,020 analyst-days of capacity."

Full platform contract: all 3 modules, full ministry/military deployment, 12-month support SLA. Annual recurring.
Revenue: $2M–$8M/yr per country
4
Year 2+
Custom Modules + Multi-Country Expansion
Each country has specific additions: Saudi Arabia wants HUMAIN interoperability. Singapore wants joint ops AI across SAF branches. Kuwait wants oil infrastructure predictive maintenance. UAE wants drone coordination AI.

Each custom module adds $500K–$2M to the contract. One country reference client unlocks the others — Gulf states all watch each other's procurement decisions closely.
Revenue: $500K–$2M per custom module per country
What You Actually Build With Claude

Six layers. All buildable by one developer with Claude as your pair programmer. No PhD required.

Layer 1 — Frontend
Chat Interface
Simple Next.js or plain HTML/JS chat UI. Secure login. Role-based views. Arabic RTL support. Runs in any modern browser.
Next.js Tailwind CSS Arabic RTL Auth.js
Layer 2 — AI Core
Claude API
claude-opus-4 for complex reasoning. claude-sonnet-4-6 for high-volume tasks like document processing and brief generation. Structured JSON outputs for all extractions.
claude-opus-4 claude-sonnet-4-6 Tool use Structured output
Layer 3 — Retrieval
RAG Pipeline
Chunk and embed all data sources. Store in pgvector (PostgreSQL extension — already available on Azure/Oracle). Retrieve relevant context before each Claude call.
pgvector Embeddings API Chunking logic Hybrid search
Layer 4 — Data Connectors
Ingestion Layer
Connectors for PostgreSQL, MySQL, Excel/CSV upload, PDF ingestion, REST API polling. Each connector is 100–200 lines of Python. Claude writes them in minutes.
Python connectors pypdf pandas REST polling SQL adapters
Layer 5 — Automation
Brief Generator
Nightly cron job reads new data, calls Claude with a structured brief prompt, generates PDF/HTML output per role, delivers via secure email or portal push.
Cron scheduler PDF generator Email delivery Role templates
Layer 6 — Deployment
Sovereign Cloud Deploy
Docker containers deployed on their existing Azure/Oracle/GCP. No special setup. Their cloud team does the deploy. You provide the Docker images and IaC templates.
Docker Terraform IaC Azure Container Apps Oracle OKE
Revenue Model Across All Six Countries

Conservative estimates. One country closed funds the next two pilots.

Phase 1 — Pilots
$1.8M
6 × $300K pilot contracts
Phase 2 — Full Platforms (Yr 1)
$12M–$30M
3–4 countries × $3M–$8M ARR
Phase 3 — Custom Modules
$8M–$18M
4 modules × 4 countries avg
Year 2 Total ARR
$20M–$48M
Fully contracted, 6 countries
Cost To Build The Full Product
6 weeks
To working MVP (you + Claude)
~$2K
Claude API cost per demo build
~$800/mo
Claude API cost per live deployment
1 dev
Team size to build initial product
The One-Line Pitch
"Give every minister and general a ChatGPT that answers questions from their own classified data — running on the cloud they already own, live in 90 days, for $300K."
Palantir charges $20M–$50M and takes 18 months to do the same thing. You're not undercutting Palantir — you're going to the 80% of the market Palantir can't serve. That market is worth $400M in Singapore and the Gulf alone.
Reports
🛰️ Project MAX — Palantir Takedown 🔍 Palantir Service Offerings 💰 Replication Targets ⚡ Sovereign Stack Deep Dive 📡 Satellite Brief 🔭 Real-Time Satellite Vision 🌍 Palantir Access Map 🟢 No Satellite Stack 🇺🇦 Ukraine Defence AI 🌏 Singapore & Gulf Analysis 🧠 The Product They Can't Refuse 🎯 Swarm Brain — Ukraine 🔮 Predictive Offensive Intel 👁️ Autonomous Border Intel 🛡️ Orb Defence — UAE 3D AI Grid