Most teams treat an RFI questionnaire as a writing task. It is really a parsing task. The workbook arrives with multiple tabs, hierarchical question sets, category metadata, and mixed response types, and someone has to turn all of that into something downstream teams can actually work with. The AWS Machine Learning Blog published a walkthrough on September 10, 2026, showing how to automate that step end to end with Amazon Quick Automate.
What the workflow actually does
The pattern is narrow and worth understanding before you copy it. A multi-tab RFI workbook sits in Amazon S3. Quick Automate reads it, extracts and structures the questionnaire data, and writes a clean CSV back to S3. The AWS post frames the pain plainly: hundreds of questionnaires a year, each one a complex workbook, and every format change forcing manual rework.
Quick Automate is described as multi-agent automation for enterprise processes that span departments, systems, UI and API interactions, and third-party systems. You describe processing goals in plain language, and it produces an executable workflow covering ingestion, transformation, validation, and output. You refine that workflow conversationally, run it against your data, and promote validated versions across AWS Regions.
The setup is mostly plumbing
Before any prompt gets written, three things have to exist. An Amazon Quick Enterprise subscription with Quick Automate access. An S3 bucket in the same AWS Region as the Quick application. And an IAM role that lets Quick Automate read and write to that bucket.
The AWS walkthrough creates a role with a custom trust policy for quicksight.amazonaws.com, then attaches permissions for s3:GetObject, s3:PutObject, and s3:ListBucket. That role ARN gets pasted into an S3 action connector in the Quick console, which is then shared with an automation group. The project itself is created under that group, with an optional business case section for tracking hours saved and target launch dates.
If you have wired up connectors before, this will feel familiar. The interesting part is what comes next.
Natural language replaces the parsing code
The prompt in the AWS example names the connector, the file, the prefix, the sheet, and the output shape. It asks for main questions and subquestions to be identified by numbering and indentation, category metadata pulled from column headers, and columns for Serial Number, Category, Question, and Response Type. The generated workflow then maps to those instructions: read data, extract and transform, structure output.
That last step is the one that matters for anyone who has done this by hand. Subquestions get converted into standalone records by merging parent context. In a spreadsheet, a subquestion is meaningless without the question above it. In a flat CSV, it has to carry that context itself.
This is the same class of problem as choosing a structured data extraction tool for AI agents: the hard part is rarely the model call, it is deciding what a record is and preserving enough context that the record survives outside its original layout.
Where the approach breaks down
The AWS post is explicit that this works without custom code for common scenarios. That qualifier is doing real work. If your questionnaires have irregular numbering, merged cells, or category metadata that lives somewhere other than column headers, the generated workflow may need refinement or may not fit at all.
The adaptation story is also worth reading carefully. The post says you can adapt to new questionnaire formats by updating the natural-language instructions. That is a genuine change from maintaining a parser, but it is not free. Someone still has to notice the format changed and rewrite the instructions, and the validation step still runs in a pre-production account before promotion.
The supplied material does not specify how the workflow handles conflicting or ambiguous question hierarchies, or what happens when a sheet’s structure deviates from the pattern the prompt describes. If your workbooks are messy in ways the example is not, test that before you commit.
A practical starting point
Pick one questionnaire format you process repeatedly. Build the connector, describe the extraction in plain language, and check whether the generated workflow produces records your downstream team would accept without cleanup. If it does, the promotion path through Import/Export gives you a way to move it into production without rebuilding it.
If it does not, you have learned something useful about your data for the cost of an afternoon.
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.