As of 13 August 2026, AI can only partly copy data between your business apps.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, power-user skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open the documentation or export-schema pages for both apps and record their object names, field names, data types, limits, authentication options and available create, update and search operations.
   2. Create fictional sample records that represent the cases the workflow must handle, including a normal record, a missing value, a possible duplicate and a changed record, and remove all customer or staff data.
   3. Decide which records may be copied, how the source and destination records will be matched, which system is authoritative, how often the copy should run and what must happen when a match is unclear.
   4. Paste the app details, fictional schemas and rules into the prompt, then ask the model for the design, mapping, permissions, implementation and test plan.
   5. Open the proposed implementation tool, such as Replit, or your existing automation platform, and add credentials through its secret-management feature rather than pasting them into the chatbot.
   6. Build the workflow with logging, duplicate protection, limited permissions and a disabled or dry-run mode, then run it only against the fictional records and compare every source value with its destination result.
   7. Change the test records to trigger missing fields, ambiguous matches, API errors and retries, confirm that the workflow stops or alerts without corrupting data, and document the rollback action before enabling real records.
   8. Enable the workflow for a small, non-critical batch, compare its logs and destination records with the source, and send the mapping and test results to a technically competent colleague before wider use.

   Prompt

   Design a safe, reversible data-copy workflow between [SOURCE APP] and [DESTINATION APP]. The source contains [RECORD TYPE], and I need these fields copied: [FIELD LIST]. The workflow should run [TRIGGER OR SCHEDULE], create or update records using [MATCHING RULE], and handle missing fields, duplicates, deleted records, API failures and retries explicitly. Use only the app documentation and sample schemas I provide below. Do not invent endpoints, field names, permissions or authentication methods. Do not ask me to paste passwords, API keys, access tokens or customer data. Mark every item that must be confirmed in the apps. Produce: a plain-English design, a field-mapping table, the minimum permissions required, implementation steps for [CHOSEN AUTOMATION TOOL OR LANGUAGE], pseudocode or configuration, a test plan using dummy records, rollback steps, logging and alerting requirements, and a list of risks. The workflow must not overwrite existing destination data unless the matching rule confirms the record, and it must fail safely when a value or match is ambiguous. Here are the source app details and documentation: [PASTE SOURCE DETAILS]. Here are the destination app details and documentation: [PASTE DESTINATION DETAILS]. Here are the sample schemas with fictional values only: [PASTE SAMPLE SCHEMAS].

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.