← All posts
·4 min read·Max Girin

What's new in the mapping wizard: clearer to read, built for real data

A round of upgrades to the platform mapping wizard: business-readable field names with plain-English explanations, a single clean sample per field, your own sample data, collection and nested-object mapping, one-click AI auto-fix with a change overlay, and undo. Easier to read — and it handles the data shapes real systems actually return.

The mapping wizard is where an integration goes from a sentence to an exact spec, so we keep sharpening it. This round had two goals: make every row readable by a businessperson, not just an engineer, and let the platform AI map the messy shapes real systems actually return. Here is what changed.

Names you can actually read

Every field now leads with a plain-English name — "First carrier terminal — city" sits front and center, while the technical path drops to a quiet second line for whoever needs it. We stripped the namespace noise and special characters out of those paths, and hovering any field explains in one sentence what it is. No code to read along.

One clean sample, not a wall of data

Each field shows a single example value right under its name, so you can see at a glance that a number is a number and a date is a date. Want to check a specific value? Open the row and type one into the test box. And you can now upload your own sample data — drop in a real record and the previews show your actual values instead of placeholders.

Transforms tucked away until you want them

The transform chips used to crowd every row. Now they live behind a per-row expand toggle, so the grid stays calm and scannable. Open a row to add rules, defaults, if/then conditions, constants, find-and-replace, or to combine fields (first + last into a full name) — all described in plain English, no code.

Built for real-world data shapes

Real records are rarely flat. The wizard now maps collections and line items (one-to-many — every order line, not just the header) and nested object sub-fields (address.city to shipping_city). Defaults are first-class, and a record-level filter lets you move only the records you mean to.

One-click AI fix — and undo

The Fix with AI button now walks the mapping, corrects the flagged issues — duplicates, type mismatches, unmapped fields — and shows an overlay of exactly what it changed, field by field. Changed your mind? Undo rolls back the last change in a click. Any field with a problem floats to the top so you never go hunting for it. Nothing runs until you approve it.

  • Business-readable field names; the technical path is secondary, explained on hover.
  • One clean sample per field — or upload your own records to preview real values.
  • Transforms behind a per-row expand: rules, defaults, if/then, combine, constants.
  • Collections / line items and nested object sub-fields, plus a record-level filter.
  • One-click AI auto-fix with a change overlay, undo, and jump-to-problem.
Readable enough for a businessperson to confirm, capable enough for the data your systems really send. Nothing runs until you approve it.

Same two-minute, self-serve step — now clearer, and far harder to get wrong. Tell us what to connect and watch the mapping draft itself.

Stop writing glue code.

Describe what you want connected. We build it, run it, and bill one flat fee.