I'd recommend using Airtable's new 'Filter record selection by a condition' toggle in the linked record field config. The AI matching uses text similarity (along with other techniques) under the surface and is not designed to deal with hard filter conditions.
One suggestion for issues with lets say the boat name where it misses stuff would be to add another field that can be updated with additional data. That way it should only happen once and then after that the alternative terms and names are there and can just be easily searched. I'm thinking in many cases the same people will use the same wording every time so over time those alternatives will be added in and a person wont need to correct it.
Do you think it could be used for a Product Recommendation Quiz? It would need to take a person's form inputs, match them to the product list, and then display the recommendation. I'm not seeing any possibility of the latter (e.g. on the form submitted page)... possibly by email but would need an integration?
I was able to replicate your AI linked record example with the Dreadnought with a similar problem using date fields, but the second Find records step returns "Found 0 Records" even though the AI found the correct matching record. The AI result reads: The matching record is "Record_name", but using either "contains" or "is" operators in the second Find records step both fail. Any suggestions?
Hey Dan how are you? What is the best way to reach out to you? We have a small project with Airtable and we would love to have you as a consultant or maybe even hands on. 🙏🏻
Airtable’s pricing model is a bitch. I don’t know how they stay in business. All this is going to do is cause people to use third-party for AI like we are already doing.
I honestly don't think they'll mind people using their own third party AI solutions. That seems like a win for everyone - larger orgs who want to deal with a single vendor and AI native features can do so, others who are priced out and have the know how can build their own solutions
I'd recommend using Airtable's new 'Filter record selection by a condition' toggle in the linked record field config. The AI matching uses text similarity (along with other techniques) under the surface and is not designed to deal with hard filter conditions.
One suggestion for issues with lets say the boat name where it misses stuff would be to add another field that can be updated with additional data. That way it should only happen once and then after that the alternative terms and names are there and can just be easily searched. I'm thinking in many cases the same people will use the same wording every time so over time those alternatives will be added in and a person wont need to correct it.
Oh wow - this is cool! We can use this!
Omg I’m stoked.
Do you think it could be used for a Product Recommendation Quiz? It would need to take a person's form inputs, match them to the product list, and then display the recommendation. I'm not seeing any possibility of the latter (e.g. on the form submitted page)... possibly by email but would need an integration?
I was able to replicate your AI linked record example with the Dreadnought with a similar problem using date fields, but the second Find records step returns "Found 0 Records" even though the AI found the correct matching record. The AI result reads: The matching record is "Record_name", but using either "contains" or "is" operators in the second Find records step both fail. Any suggestions?
Hey Dan how are you?
What is the best way to reach out to you? We have a small project with Airtable and we would love to have you as a consultant or maybe even hands on. 🙏🏻
Hey Aviv! Looks like you found our contact form and we followed up :)
Airtable’s pricing model is a bitch. I don’t know how they stay in business. All this is going to do is cause people to use third-party for AI like we are already doing.
I honestly don't think they'll mind people using their own third party AI solutions. That seems like a win for everyone - larger orgs who want to deal with a single vendor and AI native features can do so, others who are priced out and have the know how can build their own solutions
It is way to expensive 😢
@@nicohood What solutions are you looking at instead?