PREVENTABLE FUTURE / OPEN-EDI-COMPILER

Collapse the integration toll on business documents

YOUR COMPETITOR TASKS ME

A distributor gives me three expensive trading-partner integrations, their historical messages, and the people who reconcile failures. We rebuild those lanes first, then publish the compiler and runtime.

Build a compiler that turns purchase orders, invoices, acknowledgments, shipping notices, and partner-specific rules into typed mappings, validators, replay tests, and a small local runtime instead of another permanent EDI consultancy.

WHAT IT COSTS YOU

Mappings that once justified setup fees, monthly platform charges, and billable change requests become reviewable source artifacts. Your customer can change networks or maintain a partner rule without reopening a professional-services engagement.

Why the opening exists

Business-document interchange is full of standards—UBL, Peppol BIS, X12, EDIFACT—yet the money accumulates in bilateral exceptions, onboarding, validation, and opaque mapping tools. OpenPeppol demonstrates that document semantics, routing, capability discovery, and validation can be specified interoperably; many firms still pay as though each connection were a private invention.

What AI changes now

Give models the standard, partner guide, historical payloads, and reconciliation notes; have them propose a typed transform and generate adversarial fixtures. Deterministic validators and replay against accepted history decide whether the mapping is real. Every exception becomes a test instead of another consultant’s memory.

If it works

A small supplier can connect once, own its mappings, and switch transport providers. Integration work becomes finite software maintenance rather than a toll collected every time two businesses agree to exchange a document.

Why this is a plausible threat

The task rewards the same habits as my search and evidence systems: separate identity from interpretation, make every transformation inspectable, preserve failure records, and compile ambiguous source material into a bounded runtime.

What they would have to give me

One distributor with multiple partners, a lawful message archive, a domain mapper who knows why past exceptions exist, and a first release limited to a small family such as orders, acknowledgments, and invoices.

Time to regret not preventing it

Three production partner lanes in eight to twelve weeks; a maintainable open mapping system in six months.

I have not done this.