Data mapping
The agent proposes field mappings and transformations from the source and target structures. You decide the ambiguous cases, and the agent produces transformed files you can inspect before the load.
For what happens after sign-off, see Migration and cutover.
1. Agree the scope
Section titled “1. Agree the scope”Tell the agent what should move, from where to where, and what success looks like. State whether you want a plan, reviewable transformed files, or a live load.
The agent reuses the project’s connections, uploaded exports and data assessment findings, so a short instruction is enough:
Migrate Contoso vendors and open AP from the GP database to the Intacct sandbox.Produce transformed files for review first. Success: vendor count matches afterexcluding inactive vendors with no open balance, and open AP total matches to the cent.Start before you have access
Section titled “Start before you have access”You can start before anyone gives you access.
- Upload the exports from the source system.
- Upload the target’s import template, if you have one.
The agent checks the files for problems, reconciles them against each other, and produces load-ready files for the target’s own import tool. It fills the template’s columns exactly.
Missing accounts, tax settings and amounts become open questions. The agent does not fill them with defaults.
Ask for a plan only
Section titled “Ask for a plan only”To get a migration plan instead of an execution, ask for one. The agent saves the required work as project tasks, with mappings, validation, risks and cutover details in the task descriptions. It does not write to the target.
2. Review the proposed mapping
Section titled “2. Review the proposed mapping”The agent inspects the source and target structures and proposes field mappings and transformations. It resolves obvious matches itself.
The agent asks you only about ambiguous business meaning, missing required values and potential data loss. Examples are a source field with no target, or a code list that must be translated.
Review a large mapping
Section titled “Review a large mapping”For a large mapping, the agent presents the proposal as an artifact with source and target columns. You can review it field by field and mark exceptions.
Decisions you make are kept in the task description or the mapping artifact, so the reasoning survives the engagement.


3. Resolve open questions with the client
Section titled “3. Resolve open questions with the client”Some questions only the client can answer, such as which inactive customers to keep or how to treat a legacy dimension. These go into a questionnaire the agent builds as an artifact.
- Share the questionnaire by link.
- The client’s answers arrive in the project, and the agent applies them to the mapping.
- Record each decision in the mapping artifact, so the sign-off document is complete.
4. Inspect transformed files
Section titled “4. Inspect transformed files”By default, the agent produces transformed files before it touches the target. It builds a saved tool that reads the source, applies the agreed transformations and validation rules, and writes the result to an Excel or CSV file.
You can download the file from the run. Rejected records go to a separate sheet or file, with the reason for each.
- Open the files and spot-check them against the source.
- Send the rejects to the client for cleanup.
- Ask the agent to adjust the rules and rerun.
Because the transformation is a saved tool, a rerun applies exactly the same rules to the current data.


5. Sign off the mapping
Section titled “5. Sign off the mapping”When the mapping and the transformed files are right, get the client’s sign-off in one of two ways:
- Export the mapping artifact as PDF or Word for the client’s sign-off.
- Ask the agent for a sign-off form the client completes by link.
Then mark the mapping task as done. The saved transformation tool is now the agreed contract for the load.