AI-Powered Business Intelligence
AI integration platform
A Northern Michigan retail shop needed to stop pulling reports and start asking questions. We built a secure bridge between their POS and the AI tools they already use.
The problem
A Northern Michigan retail shop, and any business whose answers are locked inside a POS dashboard.
The shop's sales, inventory, customer, and service data all lived inside their point-of-sale system. Getting an answer to a simple question — last week's revenue, which products are running low, what's in the service queue — meant logging in, navigating menus, pulling a report, and sometimes exporting a spreadsheet. The data was there. It just wasn't accessible in the way people actually ask questions.
Meanwhile, the same team was already using AI assistants for other parts of their work. The gap wasn't the AI or the data — it was the connection between them.
What I built
A secure MCP server — built on the same open protocol powering AI tool use across the industry — that gives AI assistants direct, read-only access to ten categories of business data: sales summaries, inventory levels, low-stock alerts, top products, customer lookup, employee performance, transaction drill-downs, work orders, product catalog search, and serial number tracking.
The server is AI-agnostic: it works with Claude, ChatGPT, Gemini, Cursor, and autonomous agent pipelines. It's hosted on Cloudflare Workers with sub-50ms edge latency, designed multi-tenant from the start, and handles all authentication server-side so the AI never sees an API key, OAuth token, or database credential. Credentials are encrypted with AES-256-GCM at rest.
- 10 live business tools covering sales, inventory, customers, work orders, and more
- AI-agnostic — works with Claude, ChatGPT, Gemini, Cursor, and agent pipelines
- Multi-tenant with cryptographically signed onboarding
- Provider-swappable architecture — tools are named generically, not vendor-specific
- 320+ automated tests across 36 test suites with full CI quality gates
- AES-256-GCM credential encryption — the AI never sees a token or password
Why not a template
No template, plugin, or drag-and-drop builder connects a point-of-sale system to AI assistants with tenant isolation, credential encryption, and a provider-swappable architecture. This is infrastructure engineering: a four-layer system with validated tool entry points, vendor-independent business logic, integration adapters, and a shared HTTP client handling OAuth refresh, rate limiting, and retry automatically.
What it demonstrates
- 01API design and security architecture
- 02Multi-tenant SaaS with cryptographic onboarding
- 03Provider abstraction — swap Lightspeed for Shopify without changing the AI interface
- 04Production deployment on Cloudflare Workers with full CI/CD
- 05Real business value from day one, not a proof of concept
Tell me what you're trying to build.
Tell me about your business and what's getting in the way. You'll get a straight, specific reply back, usually within one business day.