What is your goal?
design a scalable, state-aware AI email workflow that can handle messy multi-step supplier conversations
What is the problem & what have you tried?
Hi everyone,
I’m curious how people here structure more complex email-reading workflows with AI inside Make.
I send a lot of price inquiry emails every month, and I want AI to read the replies, extract structured data like price, MOQ, lead time, and then create or update records in Airtable. That part is manageable when suppliers respond clearly.
But in reality, things get messy.
Sometimes they don’t reply with a clean price. They ask additional questions, send partial information, attach price lists, or respond in a long conversational way. At the same time, my inbox also receives order emails, promotions, unrelated replies, and other messages that are technically connected to the inquiry but are not what I want the workflow to process.
Then there is the multi-step conversation problem.
If AI extracts a price and then sends a follow-up question, and the supplier replies again, how do you prevent the system from creating duplicate or overlapping records in Airtable? How do you keep the workflow aware that this is still the same conversation?
How do you think about filtering, conversation tracking, record updates versus new records, and handling unclear or partially structured replies?
Would really appreciate hearing how others approach this at a higher level.
Thanks!