We recently worked with a manufacturing client who was spending hours manually entering order information from customer and suppliers emails into their system. Much of the time was spent on repetitive text cleanup and copying details across multiple fields.
Instead of adding more manual forms, we set up a Retool workflow that automatically processes incoming emails using an AI model. Here’s the flow:
- Incoming emails are saved in a database (storing subject, body, and attachments).
- A workflow triggers whenever a new email row is added.
- The email body is passed to an AI model with a structured prompt like:
- “Extract customer name, order quantity, and product code. Return valid JSON.”
- The AI response is parsed programmatically. If the AI returns invalid JSON, a quick cleanup and retry step ensures the data is correct.
- The cleaned result is inserted back into the orders table.
This approach could also work by pulling emails directly from Gmail (or other email providers) on a scheduled basis, but triggering on database insert allows near real-time processing, ideal for fast order intake.
Once implemented, employees now see pre-filled order forms that are 90% complete, only needing to verify or adjust edge cases. This reduced order entry time by roughly 50% and eliminated the need for a large external service.
Key technical takeaways:
- Workflows with retries and error handling ensure robustness.
- Validating AI output before writing to the database prevents corrupted data.
- Keeping AI prompts literal and structured avoids unpredictable responses.
- Custom prompts can outperform generic AI classification tools for specialized tasks.
Want to see how this could work for your business?
If your team is still manually entering orders or spending hours on repetitive data cleanup, this approach can save time and reduce errors.
By booking, you agree to our Privacy Policy