How to update a data flow with Coupler AI without rebuilding it

In Coupler.io, a data pipeline is called a data flow. It connects one or more sources, prepares the data, sends the resulting datasets to a destination, and refreshes them on a schedule.

You can update existing Coupler.io data flows through conversation instead of rebuilding them from scratch. Ask Coupler AI to adjust the source, destination, refresh schedule, or transformation logic while keeping the rest of the flow unchanged.

For complex dataset changes, Coupler AI will build or refine a Custom SQL transformation. The AI reads the dataset schema and writes the query. Coupler.io's analytical engine runs the query against your connected data, validates it, and shows you the result. You can review or edit the SQL before running the updated flow.

Updating a data flow changes how Coupler.io imports, prepares, or sends your data. It does not edit records in the original source application.

What you can update with AI

Use the Coupler AI Agent inside the app or a supported AI tool connected to Coupler.io to maintain an existing data flow.

You can ask AI to:

• Update a source - change the entity or report, date range, filters, fields, or another supported source setting.

• Add another source - bring a new dataset into the existing flow for blending or comparison.

• Update a transformation - modify the data set with filters, data blending, aggregations, new columns, or custom SQL.

• Update a destination - change the target spreadsheet, sheet, dataset, table, write mode, or another supported destination setting.

• Add a destination - send an existing dataset to another supported destination without rebuilding the source setup.

• Change the refresh schedule - enable, disable, pause, or resume automatic refreshes; change the interval, days, start time, or timezone.

• Rename the data flow - give the existing flow a clearer name from the chat.

• Run the data flow - fetch the latest data after Coupler.io confirms that the flow is fully configured.

• Inspect the setup - check the sources, datasets, schema, destination, last successful run, row count, and other available metadata.

• Update dataset context - document metric definitions, business rules, column meanings, synonyms, and known data issues for future AI conversations.

How to update an existing data flow with Coupler AI

Use the Coupler AI Agent or a connected AI tool to start a conversation

• Request an inspection of the current data flow setup. Ask for a summary of the sources, datasets, destination, and refresh schedule.

Example:

Review the data flow called "Cross-channel marketing report." List its sources, datasets, destination and refresh schedule. Do not change anything yet.

This helps you confirm that the assistant is working with the correct flow.

• Describe the change and what must stay unchanged. State the desired result, not only the feature you want to use. Include the relevant dataset, fields, conditions, and business rules. Also name any settings the assistant must preserve.

Example:

Update the existing transformation. Add LinkedIn Ads to the combined marketing dataset, match its fields to the Google Ads and Meta Ads fields, remove duplicate campaign rows and calculate ROAS as revenue divided by spend. Keep the current destination and refresh schedule unchanged.

For a complex update, ask for 1 main change at a time. You can refine the result with follow-up messages.

• Review the proposed setup or SQL. Check the parts affected by the request:

  • The selected source account, entity, report, date range, filters and fields.

  • The dataset names, join keys and column mappings.

  • The generated SQL query and resulting schema.

  • The destination, target location and write mode.

  • The refresh interval, days, time and timezone.

If something is unclear, ask Coupler AI to explain the choice or revise it before you run the update.

• Apply the update. Confirm the change when the setup is correct. If your request could have more than 1 interpretation, the assistant may ask you to choose an account, dataset, field, destination, or another required option.

• Run the data flow. Ask AI to run the updated flow. Coupler AI checks that the flow is fully configured before starting the run.

Example:

Run the updated data flow and tell me whether it completed successfully.

• Check the resulting data set.

Review the output before using it in a report or analysis. Check the row count, column names, data types, key totals, duplicate records, and a sample of the results.

Example:

Check the updated dataset. Confirm that the expected columns are present, find duplicate date and campaign ID combinations and compare total spend by source.

• Update the data set context if its meaning changed.

If you added a metric, renamed a field, changed a calculation, or introduced a new business rule, ask the assistant to update the dataset context. This makes the definition available in future conversations.

Example:

Document that ROAS in this dataset means attributed revenue divided by ad spend. Note that duplicate campaign rows are resolved by keeping the latest record for each date and campaign ID.

What still requires your review

Coupler AI helps you configure and maintain the flow, but you remain responsible for the final setup and result. Review the following before using the updated data:

• Business definitions - confirm how metrics, segments and attribution rules should work.

• Join keys - make sure the selected fields connect the correct records.

• Deduplication rules - decide which duplicate record should be kept.

• SQL logic - review the generated query, especially for financial or other high-impact reporting.

• Output schema - check whether downstream reports depend on column names, order, or data types.

• Destination settings - confirm the target and write mode to avoid replacing the wrong data.

• Schedule settings - confirm the timezone and refresh interval.

• Results - compare key totals, row counts, and sample records with the expected output.