Map source data to the schema your system expects.
EDI Platform · TransformsA Transform is a reusable rule that turns a field from a source document into the shape your target system expects. Three layers cover everything from a 1:1 field copy to deeply nested arrays, with AI-generated code based on your real documents.
Why hand-rolled mappings stop scaling at partner #20
Per script, per dev
Each new supplier means a new transform script - written, reviewed, and maintained by an engineer who knows that partner's quirks.
Schema drift
A partner adds a field, renames an attribute, or flips a unit - and the mapping fails downstream without a clear error.
No shared rules
Partner-specific weirdness leaks into the global rules, or every partner ends up with their own copy-pasted fork.
How a transform gets built
Define the target schema once
Step 1Tell Chainfill what your ERP expects - field names, types, required vs optional, nested item lines. One schema, reused across every supplier.
AI proposes the mapping from real docs
Step 2Chainfill reads the supplier's actual files and proposes a transform - direct copies, concatenations, splits, array iterations, coercions, null handling - all visible in a diff view.
Review, tweak, ship
Step 3Approve the proposal, override any field by hand, or refine in the AI assistant. Calibration learns from every correction so the next supplier comes together faster.
Promote across scopes
Step 4Keep the partner-specific quirks at the supplier scope; promote anything reusable up to the company scope so future suppliers inherit it automatically.
Three layers, one engine - visible in the editor
Direct mapping, concatenation & splitting, array-aware paths - every transform sits in the same editor with the AI's proposal, your overrides, and a one-click ship.
Transforms · Order intake
Mapped from ACME EDIFACT D.96A → ERP order schema
What AI-generated transforms actually save
Days to weeks
Engineer reads partner spec, writes mapping code, handles edge cases, ships, then maintains it forever.
Reviewed mapping
AI proposes the mapping from real documents. Ops reviews, approves, tests against the partner's real schema, and ships.
Quality compounds across suppliers
Every correction trains the AI for the next mapping - so onboarding partner #50 is faster than partner #5, not the other way around.
