Transferring data between programs can reduce repeated typing, but requires knowing the meaning of the fields. A technically successful link can produce incorrect information if the two applications interpret quantities, dates or codes differently. Verification must relate to the content, in addition to the transfer.
I propose that we start with a circumscribed process and a few representative cases. It is necessary to define which system produces the information and which uses it. The responsibility for the correction must be understandable, preventing two copies from being updated separately without a shared criterion.
We imagine an exported order with an amount expressed in packs and imported as individual pieces. Rows may appear complete, but the next job uses a different data. A comparison with real examples makes this difference visible before the extension of the link.
Verification includes exceptions and updates. A useful transfer must make it possible to recognize what has not been processed correctly. The expected result is a more reliable passage of information, with a preserved meaning and a clear path when something requires a person's intervention.