Pipelines

AI in Pipelines

Pipelines use AI in two places. Both leave the final say with you, and both count toward your organization’s monthly AI allowance.

1. AI rename: suggested column names

The AI reads your source columns, plus any documentation you provide, and suggests clearer destination names. Cryptic names such as CST_ID or ORD_DT_TM are where it helps most.

Where to start it

  • When you create or re-run a Pipeline: tick Run AI Rename on the Documentation step.
  • Any time later: in the mapping editor, open the Documentation tab and click Run AI Rename. This needs text in Source System Documentation.

What you give it

  • Source System Documentation: data dictionaries, column descriptions, contracts.
  • Additional Instructions: your naming rules, for example Use Customer instead of Client.

What you get back

The rename runs in the background, table by table. From the Documentation tab you watch a Running AI Rename dialog; after create or re-run it runs after the editor opens. Suggestions then appear under each destination column, for example Suggest: CustomerId (92%), with Accept and Reject.

  • Accepted names show in blue. They are kept when you re-run the Pipeline.
  • Names you type yourself show in bold and are never replaced by a suggestion.
  • The Summary tab counts suggestions generated, accepted, and pending. Validation warns while any are unanswered.

See Map source columns to target columns.

2. Expression check for the target dialect

When a calculated column is produced through a helper view (Materialize via view), Kenseme asks the AI to check the expression against the destination database when it generates the load. If the expression will not run there, the AI can supply a corrected version, and the load uses it.

The result shows as an icon in the column’s Strategy cell: validated, rewritten (hover to read the new form), or a warning. A result is reused until the expression or the destination dialect changes.

Override expressions are not checked. They are used exactly as you wrote them. See Override a mapping with an expression.

What is not AI

  • Analyze All on the destination schema’s Pipeline tab recommends load strategies from column patterns and relationships. It is rule-based.
  • Naming conventions, validation, finalize, and code generation are deterministic.

When your AI allowance runs out

  • Creating or re-running a Pipeline still works. The AI rename is skipped.
  • Run AI Rename on the Documentation tab stops with Monthly AI usage allowance exhausted — buy a usage pack or upgrade your plan.

See AI token allowance and usage.