Airtable announced an upcoming AI field type which could generate images, summarize text etc. From the outside, the vision here is to be able to leverage OpenAI within fields themselves. You configure which field you want to (take these fields, generate an image) and the AI field generates it. My guess would be its using OpenAI in the backend and when you configure the field, you’re basically configuring API parameters. That's what we see from the screenshot (shown below).
At first glance, it makes sense. But the more I thought about it, the more reservations I had.
AI is not a field type, it's an output.
The first thing I tell to people learning Airtable is that you have to define what you're putting into the field when you create it. Dates are inputted into a date field, text into text fields etc. This is counter intuitive to users coming from spreadsheets where you have cells created and field types are defined later.
An AI field breaks one of the core principles of Airtable: field types map to data types. The AI field however can create different types of information. It can generate an image, text, a formula anything really. Obviously to maintain consistency, users can simply define the output at the beginning similar to how you create a formula field!
So the flow becomes users define a data type, set inputs for the field (e.g. summarize these other fields, or take prompt from this other field), and then the field generates a value when those fields are filled out.
Should Airtable AI live in automations?
Doesn’t that sound like an automation?
Why not leverage existing functionality? Case in point, I’ve already used the OpenAI API in Airtable to create short descriptions for my blog post with a short Airtable script in an Airtable automation. A native OpenAI action would make this type of workflow much more accessible. It would also allow for a future where Airtable would like to offer multiple AI providers to users and a wealth of parameters.
Now, putting it in Automations has real usage tradeoffs: few users get to the level of sophistication in Airtable where they leverage Automations. By making it a field, Airtable is putting AI in front of many more users. This approach means using AI is as easy as adding a date field! You can generate images in seconds! That'll certainly get more people to use it!
That ease of use comes at the cost of flexibility. You're out of luck if your usecase doesn't fit neatly into what Airtable predicts you want to do with AI.
Does making it a field mean that overall people will have better workflows? Get more value out of Airtable? Or is this a case where the race of getting AI out as quickly as possible has won over the long term value it can bring? We'll wait and see but as a power user, I worry about this decision.