Ask about a specific job
A migration starts with questions hidden inside the job: why a row is rejected, how an amount is calculated, or which condition starts the next activity. Open Ask AI Migration Assistant from the Catalog or job inspector and enable job context. The conversation stays attached to that project and job, including follow-up questions.
Ask for an explanation of a stage, column or expression using the imported design. That gives an engineer a concrete starting point for understanding business rules and planning their target implementation.
Go back to the source
The AI Migration Assistant can use stage names, links, SQL, derivations and schemas to explain the source-to-target flow. Answers can link back to the supplied job and saved migration evidence so you can inspect the material behind the explanation.
Job context is read again for each question. For a large job, name the stage, column or file you want to discuss: relevant stages are prioritized and some details may be excerpted or omitted.
Keep acceptance with the team
Ask the AI Migration Assistant to suggest review questions and test cases from the available logic. Use those suggestions to investigate null handling, joins, filters, reject paths and differences in target behavior.
Your team reviews the code, runs the workload and reconciles results. The assistant provides guidance; it does not execute jobs, query live databases, change migration records or approve a migration.
Start the conversation from the job
Open Ask AI Migration Assistant from a job row or the job inspector to bind a conversation to that exact project and job. Preview the source context, then choose Use job context to authorize sharing selected evidence with the configured provider for that conversation.
The context can include stages, links, SQL, derivations, schemas and sequence activities. Where migration records exist, it can also include recorded checks and revision information, with bounded saved-code excerpts for code questions.
Keep provider choice and context visible
The assistant uses the signed-in user’s configured OpenAI or Anthropic connection. General workspace sharing is off by default; job-context consent applies separately to the selected conversation. Context links and source fingerprints help you return to the supporting job.
Evidence can be excerpted or omitted to fit context limits. The assistant explains and investigates the available material; it does not execute jobs, query live databases or edit migration records. Use the code-generation and review workflow to prepare migration artifacts.
What the AI Migration Assistant can read
With job context enabled, selected evidence from your imported DataStage job is shared with your configured AI provider. The available fields depend on what the import captured and which migration artifacts have been saved.
| Available context | What you can investigate |
|---|---|
| Job design and connections | Job description, stage names and types, connected stages and link labels. Ask for a walkthrough from inputs to outputs. |
| SQL and table references | Imported SQL, table names and write modes. Ask which data is selected and how the job writes its result. |
| Transformation rules | Column derivations, stage variables, loop variables, constraints and reject settings. Ask how a business rule is calculated or why a row takes a particular path. |
| Column schemas | Names, types, nullability, precision, scale and length where captured. Ask which type and null-handling cases need target tests. |
| Operator behavior | Selected join, filter, grouping, aggregation, sort and duplicate-handling settings. Ask what behavior needs to be preserved during conversion. |
| Parameters and sequences | Parameter names and types; sequence activities, child-job references, expressions, triggers and loop conditions. Ask about execution conditions; a child-job reference does not include that child’s full design. |
| Saved migration evidence | Destination, revision metadata, recorded checks, code-review decisions and notes, plus saved artifact details and blockers when available. Ask which recorded findings deserve attention. |
| Saved target code | For code-related questions, bounded excerpts from saved migration files. Ask how the available code represents a source rule and what needs closer review. |
Use it throughout the migration
Understand the existing job
Ask: “Walk through the transformation rules and filters in this job.” Use the explanation to prepare a source review and identify business logic that needs an owner’s confirmation.
Plan the target implementation
Ask: “What should we review when moving this job to Fabric?” Discuss how the supplied stage logic could be implemented on Fabric, Databricks or Snowflake. Treat proposed mappings and code examples as guidance to review against the target runtime.
Review the saved conversion
Ask: “Compare the saved target code with this job’s derivations. Which rules need closer review?” Where artifacts exist, the assistant can discuss the source and saved code excerpts together. Use the conversion workflow to generate and save the actual package.
Prepare validation and handover
Ask: “Which tests would exercise this job’s nulls, joins and reject paths?” Turn the suggestions into a validation checklist, then review the recorded checks and unresolved questions with the receiving team. Saved checks do not independently establish current execution or equivalence.
Choose what you share
Configure your own OpenAI or Anthropic provider connection in AI settings. Starting a job conversation makes no provider request. Preview the context and choose Use job context before sending selected evidence for that conversation.
Context comes from imported metadata and saved artifacts. Raw XML, connection settings and parameter values are excluded; recognized credential patterns are redacted. SQL and code can still contain confidential business literals, so review the source before sharing.
General workspace sharing is a separate setting, off by default. Its opt-in supplies matching workspace names, not complete job logic. Select a job and enable its context for detailed questions about that job.