Skip to main content
Most of the work of onboarding a customer is converting their messy export into your clean schema. Vern’s migration agent does that for you. It reads the files, works out how they map onto your templates, writes a script (the recipe) to do it, previews the result, and asks you whenever something needs a decision. Nothing is written to the workbook until you click Run import.

​
Start an import

Open the company and click Import (or Re-import from the workbook’s ⋯ menu). Then on the upload screen:
  1. Add the files. Click or drop files into the upload area.
    • Accepted: CSV, Excel (.xlsx, .xls, .xlsm, .xltx, .xltm), SQL dumps (.sql, .bak), XML and ZIP archives of any of these.
    • Up to 5 GB per file, 20 GB per upload, and 400 files per import.
    • Files already in the workbook from an earlier import can be brought in too — turn on Include previous data.
  2. Add instructions (optional). Instructions for this migration is for anything specific to this run — “ignore archived jobs”, “phone numbers are all Australian”.
  3. Click Start import.
If the company’s source has an API connector, you can switch on Extract via source name instead of uploading — enter the customer’s credentials once and click Extract (or Extract & upload to combine both).
PDF, Word and image files (scanned documents) are temporarily unavailable. Uploading one shows “Document extraction (PDF, image and presentation files) is temporarily unavailable. CSV and Excel files are supported.” Documents inside a ZIP are skipped and listed as skipped.
If you upload a file whose name matches one already in the workbook, Vern asks whether to Skip it, Upload anyway, or Replace the existing one.

​
Watch the agent work

The import opens as a console with four tabs: Along the way the agent may stop and ask:
  • Questions — how to read an ambiguous column, what to do with partial records, which of two date formats is right. Pick an answer or type your own, then click Continue.
  • Template updates — when the data has something your templates can’t hold, it proposes the change. Click Apply updates to accept, or carry on without. Turn on Template lock on the company to stop it proposing these.
  • Conflicts — when the data contradicts a template rule, it shows both sides and asks which wins. Click Apply on your choice.
For things it can look up itself — what a source system’s status code means, an industry abbreviation — the agent searches the web rather than asking you.

​
Steer it

There’s one message box at the bottom:
  • Type a change and click Apply edit — “split @Name into first and last”, “drop rows where status is test”. The recipe and preview update.
  • Switch on Ask to ask a read-only question — “what happens to rows with no email?” — without changing anything.
  • On any of your messages, click Undo from here to roll the import back to that point.
  • The ⋯ menu picks the Model — Speed, or Thinking for hard migrations — and has Clear chat.

​
Run it

When the preview looks right, click Run import. The agent runs the recipe over all the data and writes the rows to the workbook. When it finishes you’ll see Successfully imported N rows, and the sheets show every value checked against your template rules: invalid cells are tinted red (often with an AI-suggested correction to pick from), and warnings from AI rules are tinted orange. Calculated and extracted columns are filled in as part of the run.

​
Stop or cancel

  • Stop ends what the agent is doing right now. It stops on the server, not just in your browser.
  • Cancel import abandons the run and returns you to the upload screen, keeping your files and answers.
Only one import can run for a workbook at a time.

​
Imports get faster

Every import produces a recipe — the portable logic for “this source → your templates” — and keeps the customer-specific details separate. On the next import from the same source system, the agent starts from the best earlier recipe instead of from scratch, so the second customer you onboard from a system is much faster than the first. To make one recipe the default for a source, use Publish & bind import recipe on the source. What the agent learns about the customer goes into the company’s Knowledge tab, and what it learns about the source system into the source’s knowledge — both are read on every later run. Turn off Memory on the company to keep a test run from saving anything.

​
After import

If something needs fixing — a value didn’t normalize, a rule was misread — spot-check and fix in the sheet.

​
Next