Airtable AI Is Here 🤖 🪄

Summary

This video introduces new AI features in Airtable, focusing on automation improvements and AI-generated text. Dan Leman from Automation Helpers explains how AI can link communication records to ship names in logistics, streamline workflows, generate tweets, create formulas, categorize information, and recommend linked records. The video highlights the potential of these features to save time and improve accuracy. It also discusses pricing, usage, and how to access these new capabilities. Dan encourages viewers to explore Airtable AI and offers free consultations for further assistance.

Takeaways

• 🚀 Airtable AI is now generally available, featuring exciting new automation capabilities.
• 📝 Generate with AI can take inputs from previous steps and use them in subsequent steps for enhanced automation.
• 🚢 Example use case: Managing logistics for a large fleet of shipping vessels using AI to link communications with ship names.
• 🔍 The process involves retrieving a list of ships, generating AI prompts, and finding the closest match to link records.
• ⚙️ AI in Airtable automations can streamline workflows by integrating inputs and outputs across multiple steps.
• 📄 AI features can be used for various text generation tasks, such as creating tweets, press releases, and LinkedIn posts.
• 📊 AI capabilities extend to categorizing information and performing sentiment analysis within Airtable.
• 🧮 AI can help generate formulas by interpreting prompts and creating appropriate functions within the database.
• 💼 AI can recommend linked records based on fields like region and role, improving task assignments and project management.
• 💸 Airtable AI is available as a paid feature with limited-time free credits, requiring a subscription for continued use.

Q & A

What are the new features introduced in Airtable?

The new features introduced in Airtable include AI integration for generating text, automation enhancements, and the ability to link records based on AI-generated matches.

What is the significance of the 'Generate with AI' action in Airtable automations?

The 'Generate with AI' action allows users to take inputs from previous steps and use the AI-generated output in subsequent steps, enhancing the automation capabilities in Airtable.

Can you provide a use case example where AI in Airtable automations is beneficial?

A use case example is managing logistics for a large fleet of shipping vessels, where AI helps link communication records with ship names even if they are not spelled correctly.

How does the AI feature handle ship names that are not spelled correctly?

The AI feature can match the closest ship name from a list provided, even if the ship name in the communication is not spelled correctly.

What limitations did the speaker encounter with the AI-generated output in Airtable?

The speaker noted that the AI could not return an Airtable record ID directly, which required an additional step to find the record by name.

How does Airtable ensure the accuracy of AI-generated matches for linked records?

Airtable uses conditional logic to proceed only if there is exactly one match found. If there are multiple matches or no matches, it updates the record with a note instead of proceeding.

What other surfaces can AI be used in within Airtable besides automations?

AI can be used in generating text for long text fields, creating tweets, press releases, LinkedIn posts, translations, and categorizing information such as sentiment analysis.

How does the AI feature in Airtable handle inline record creation and updates?

The AI feature in Airtable does not query every keystroke when creating inline records. Instead, it waits for the user to click the 'Generate' button to update the record with AI-generated text.

How can AI assist in generating formulas in Airtable?

Users can prompt AI to generate a formula based on a description. The AI interprets the prompt and creates the corresponding formula field, even handling minor typos.

What is the pricing structure for Airtable AI features?

Airtable AI features are available for $6 per seat per month on the annual plan or $7 on the monthly plan. Each seat gets 3500 monthly AI credits. Additional credits can be purchased if needed.

Outlines

00:00

🚀 Exciting New Features in Airtable Automations

Dan Leman introduces the new features in Airtable AI, highlighting the ability to generate text with AI and its integration into automations. He expresses excitement about the potential of these features, particularly in logistics management for shipping vessels. Leman explains how the AI can match ship names from communication logs with records in Airtable, using examples and a step-by-step breakdown of the automation process.

05:00

📊 Leveraging AI for Various Tasks in Airtable

Leman discusses additional uses of AI in Airtable, such as generating tweets and other text fields, emphasizing the flexibility of AI across different surfaces. He describes how AI can take inputs from long text fields and generate appropriate content, providing detailed instructions and examples. He also mentions the capability to customize prompts and templates to streamline repetitive tasks.

10:59

🔍 Sentiment Analysis and Formula Generation with AI

This section focuses on using AI for sentiment analysis and creating formulas in Airtable. Leman explains how AI can categorize and prioritize tasks based on sentiment and feedback. He demonstrates generating a formula field using AI and discusses the efficiency and accuracy of AI in understanding and implementing user prompts, even with minor errors or typos.

16:01

👥 AI-Assisted Record Matching and Usage Limitations

Leman explores the use of AI for recommending linked records in project management, highlighting the importance of sufficient training data for accuracy. He shares his experience with AI recommendations and addresses potential limitations. Additionally, Leman outlines the credit-based usage system for Airtable AI, explaining the cost and availability of AI features based on different subscription plans, and provides insights into managing AI credit usage.

Keywords

💡Airtable Automations

Airtable Automations refer to the feature within Airtable that allows users to automate repetitive tasks by creating workflows that trigger actions based on specific conditions. In the video, Dan Leman discusses the new AI capabilities in Airtable Automations that enhance workflow efficiency, such as linking communication records with ship names.

💡Generate with AI

Generate with AI is a feature in Airtable Automations that enables the generation of text based on AI inputs. This feature can take inputs from previous automation steps and use them in subsequent steps. Dan illustrates this by showing how AI can match ship names from various communication sources, despite spelling variations.

💡Linked Records

Linked Records in Airtable allow users to create relationships between different tables by linking fields. The video explains how AI helps link communication records to ship names in the logistics management use case, streamlining the process of associating messages with the correct ship.

💡Trigger

A Trigger in Airtable Automations initiates an automation workflow when a specified event occurs, such as a form submission. Dan describes using a form submission as a trigger to start the process of matching ship names with communication records using AI.

💡Find Records

Find Records is an action in Airtable Automations that retrieves data from a table based on specified conditions. Dan demonstrates using the Find Records action to gather a list of ships, which is then used by AI to find the closest match to a referenced ship name in communications.

💡Conditional Logic

Conditional Logic in Airtable Automations allows workflows to make decisions based on specific conditions. Dan highlights the use of conditional logic to handle cases where the AI finds one, multiple, or no matching ship names, ensuring the automation behaves correctly in each scenario.

💡Sentiment Analysis

Sentiment Analysis is an AI feature that evaluates text to determine its emotional tone, such as positive, negative, or neutral. The video discusses how Airtable's AI can be used to categorize feedback or comments, helping teams prioritize tasks based on sentiment.

💡Formula Generation

Formula Generation with AI in Airtable allows users to create formulas by providing natural language prompts. Dan shows how this feature can simplify the process of writing complex formulas, like calculating the difference in days between two dates, even if the prompt includes typos.

💡AI Credits

AI Credits are the units used to measure the consumption of AI-powered features in Airtable. The video mentions that users get a limited number of free AI credits initially, and additional credits can be purchased as needed, with different pricing tiers for various usage levels.

💡Template

Templates in Airtable AI are predefined structures for generating specific types of text, such as tweets, press releases, or LinkedIn posts. Dan explains how using templates can streamline text generation by providing a framework that incorporates necessary formatting and character limits.

Highlights

New features in Airtable automations are now generally available.

Dan Leman from Automation Helpers.com discusses the benefits of new Airtable automation features.

Ability to generate text with AI in Airtable, useful for many applications.

AI can take inputs from previous steps and use the outputs in subsequent automation steps.

Example use case: managing logistics for a fleet of shipping vessels by linking communication records to ship names using AI.

AI helps in matching ship names in communications with existing ship records, saving time for the team.

AI features can generate text fields in Airtable, which can be customized and structured to avoid repetitive prompting.

AI can categorize information, perform sentiment analysis, and recommend linked records based on dynamic data.

Generating formulas with AI can help create complex formulas easily in Airtable.

AI can recommend appropriate linked records, improving task assignments in project management.

Limited-time free offering of Airtable AI credits, with pricing details for continued use.

Usage examples and limitations of AI features, including credit consumption and practical applications.

AI features in Airtable can enhance various workflows, providing significant time savings and accuracy improvements.

Enabling AI features in Airtable requires team or business plan, with admin panel controls for credit usage.

Additional AI credits can be purchased in bulk, with pricing tiers for different usage levels.

Transcripts

air table AI just HD General

availability meaning you can now access

these new features we were really eager

to talk about this for a long time but

the terms of the Clos beta said we

couldn't talk about it so now we're

going to dig into all the really

exciting new features hi I'm Dan Leman

from Automation helpers.com and we help

companies get automated with portals

apps and Integrations now all the big

YouTubers would say save the most

exciting parts for the end but I can't

help it I'm too giddy about these

features I'm really pumped about what's

happening with air table automations so

I want to talk about this first some of

the other things like generating text

with AI I'm like eh okay you know it's

going to be helpful for somebody but I

really think there's so much value in

what can be done with automations so the

first thing I want to talk about is when

you're looking at your automations and

you see an action for Generate with AI

it's not really that exciting looking I

mean if you're used to chat GPT you're

like okay well there's a spot to put in

a prompt and a couple settings about the

model itself so what's the big deal

about this well the part that makes us

so powerful is that we can take inputs

from previous steps like we would with

any other automation step and we can

take the output and we can use that in

additional steps later on so let's walk

through a use case one of our clients

manages logistics for a large Fleet of

shipping vessels and these ships all

have different names now I just grabbed

a random list of ship names cuz I don't

know anything about ship names but we

put them here because what was happening

is they have a log of communications

coming in and these communications we're

writing from different sources we've got

from twilio we've got calls that are

coming in there could be text messages

there could be emails and in these

communications they're referencing the

names of these ships so in a perfect

world we're writing in these

communications we'd be able to link the

communication records with the ships

that we have but you can imagine the

people who are working on these vessels

aren't actually writing out the full

name or talking about the full name of

the ship so we've got the HMS

dreadnotmusic linked relationship now

you can of course do some sort of

regular Expressions to try to make stuff

work out but when the ships have

different names that becomes a little

bit tricky so let's go ahead and see

this in action the actual results and

then we'll dig into the automation

itself I'm just going to fill out a form

but again this could come in from a

variety of different data sources so

this of course is a terrible example but

the dread knot is lost at C let's go

ahead and submit our form and back in

air table we can see our automation has

just updated this we have our record and

it's updated this linked record to to

our ship so now we can see the name of

the HMS dreadnots able to find that

linked record and create that

relationship between both of those

records so how did we do this in the

automation itself okay so we have a

trigger really standard we're just

saying when a form's submitted we could

use a different trigger if we want now

the first step that we do the first

action here might appear to be a little

bit counterintuitive the first thing

that we want to do is to find records

and we want to get all of the different

ships that we have it's pretty rare that

you're asking for all of the records

inside of air table normally you're

adding some sort of condition to this in

this case we want to retrieve the list

of all of the ships because we want to

use that to feed into our prompt to say

hey you've got this list in front of you

which is the closest one to what you're

looking for so now we have our

generative step and here's where we put

in our prompt I just put in a

description I said okay you've got a

body of text and a ship name is being

mentioned the ship might not be spelled

the same way but your job is to find the

closest match name and return the name

and then the body of text is and this is

where I'm able to pull in the attributes

from the record so I fed it the notes

the dread knot is lost at C and then we

say please find the closest ship name

here's your list of ships to choose from

and this is where we're grabbing the

list of records that we retrieved in the

previous step now I did a little bit of

testing here and I really wanted it to

find the Matched ship and then be able

to return the record ID so we wouldn't

have to do an extra step here now un

unfortunately it kind of gave me the

standard chat GPT response like we've

only been trained on data up to this

point and we can't give you an air table

record ID so maybe it's possible maybe

there's a different way to do it I

wasn't able to get it to do it but in

this case this wasn't a problem because

it was able to return the exact name

since the exact name we retrieved here

and then all we need to do is another

find record step so now we can say where

the name contains and we can plug in

that AI response so from our previous

step we say if there's a match on the

name that's the ship that we're going to

return and then we've got a little bit

of conditional logic here to say if the

records length equals one that means if

you were able to find a ship do this

step if you couldn't find a match don't

do it if you found multiple matches

which we don't want don't do it but if

you found exactly one match then we want

to proceed we're going to update our

record and we're going to then link this

to the corresponding ship that we

retrieved the record in that step

otherwise if we didn't find a match or

we found multiple matches then we're

going to update the record and we're

just going to add a note and say we

couldn't find the Matched ship via Ai

and of course as we all know AI isn't

perfect so there's going to be a number

of times where it doesn't find that

exact match but even if it could help us

out 60% of the time that would still

save a lot of time for the people who

are triaging these requests coming in so

I hope you're as excited as I am about

all the possibilities with air tables AI

in automations especially because we can

take inputs plug it into our prompts we

can take the response and do other

things with it afterwards so I think

this really opens up a lot of new

possibilities okay so we've talked about

how we can use AI in automations where

else can we use it in the application we

take a look at the support article it

talks about different surfaces where we

can use it and surface is kind of an

interesting name so in air table they

can't say What fields can we use this in

because they also work in automations

and we can't say what automations cuz it

also

generate text which is going to give us

these additional options here now you

can click and edit these properties in

this particular one we're generating a

tweet and this is a template that

already knows in the background what a

tweet is comprised of I'm surprised

they're using the word tweet what's the

word now is it X XZ I don't know what it

is but anyways generating a tweet it

probably already knows how many

characters that you have available to it

maybe information about adding hashtags

and so really that information is

already kind of in the background you

don't have to worry about it cuz it's

baked into that template so the most

important thing here is identifying

which field we're pulling from for our

prompt information so we've got this

content description you fill that out

that's what's going to give it the

information it needs to be able to

create the text here we can also include

additional fields that we want and so

you can select those and it will

regenerate based on the information that

you give it you can add your own

examples so if you add certain kinds of

tweets and you want to mimic that style

you could do that here and you can give

additional instructions so really what

this is doing is it's making it so that

unlike chat GPT where you give it some

information it comes back you add to

your prompt it comes back so instead of

doing that you can avoid all of that

because you're giving it that structure

kind of that metadata that lives behind

the scenes so that you don't need to

reprompt every time this in effect is

essentially acting like a custom GPT

you're giving it all these instructions

and then it's taking the dynamic data

The Prompt that you have in this content

description to use that to generate the

text now now like we said the long text

Fields have the majority of AI features

attached to it to generate different

kinds of text there's some different

examples we have here like a press

release and Linkedin post and things

like that you can create a new field and

in this you'll see this new option to

create an AI field well it's not really

an AI field there's not a new field type

here really we're just choosing from

existing things so either you could

create the long text field and then

enable those AI settings or you could

simply click here and then you could say

let's generate text which is then going

to create that long text field so it's

really up to you whichever is more

convenient once this is toggled on you

can choose from either starting from a

template or starting from scratch if you

start with a template you've got some

different options here that are

categorized General product marketing ux

research and recruiting I'm always a

little bit surprised about these

categories and they're doing this for

the apps too like ux research is really

specific in my mind but that's a

different story so maybe you want to do

a translation this is one of the use

cases they feature quite a bit here we

can click on that you can choose what

field you want to translate and then the

language you want to translate it into

I'm kind of surprised this isn't a

dropdown that you just put in free text

what language it is so I'm kind of

curious why that is and again you could

give it additional instructions so

pretty straightforward to use long text

fields for the purpose of text

generation next we're going to take a

look at it through the lens of being

able to categorize information so

there's another base that air table

created to help walk through some use

cases what's interesting here is for

something like sentiment analysis if we

edit this

with inside of air table now if you're

creating a record in line I was kind of

curious how this would handle it because

air table is so convenient to allow us

to create inline records but what it's

doing is it creates the record right

away followed by a bunch of updates and

given that you only have certain limits

around the amount of consumption you

have of the AI features I was like oh

man is it going to make that query every

single time that you're updating that

record thankfully it's not so if we

create this new record in line we see

these little text messages here we can

expand this you can see that it says

okay generate text in order for this

field to work and so I don't want it to

be every single key stroke that I'm

typing in the feedback so if I were to

do it in line and I'm adding the

keystroke it's not going to make that

query every single time notice that it

now has that button for Generate so now

if I were to update this value with a

message and I click off this has

technically run an update but I still

need to press the button either in line

here or I could expand the record to

generate that text I think this is a

good way to implement it because we're

not suddenly racking up the bill so

quickly having it update every single

time so if I want the sentiment analysis

from this I can click generate negative

our priority we can click on that and

I'm always kind of interested you know

how do they determine priority so in

this case it was high and back in our

prompt we're able to see it summarizes

high are critical issues affecting core

functionality or significant user base

so I don't know how it actually know if

it affects a significant user base it's

not actually looking at the body of

Records it's only looking at that

individual record so I don't think it

would actually have the knowledge to

know that but in this case because I

can't log in and it's crashing it's

determined this is high priority now

another thing you can do if you're

actively using the comment section in

your records with your team is if you

have a whole stream of comments and you

want to summarize it there's a little

button to be able to summarize the

comments and display this information at

the top of the screen so here you can

see a little bit of summary of what's

going on we want to bring in a partner

to work on this and here's our standard

Bill rate so it's able to summarize the

different comments that we have in that

thread next let's talk about creating a

formula with AI I'm just going to do a

simple one here we'll insert a field and

we'll choose our formula field like

normal and you'll see this option to

generate a formula so I'm just going to

give it a prompt here to find the

difference in the number of days between

the start and the deadline those are two

different fields that I have here let's

go ahead and generate and you can see it

now created our datetime diff function

show us the number of days let's go

ahead and create that field and we can

now see the number of days what's cool

is that I used lowercase and it was

still able to interpret it correctly I

didn't have to put that field name

exactly as is let's try actually with a

typo just to see if we mistype it so I'm

going to delete this let's copy the

description that we generated we'll

click on generate formula again paste

this in and let's just make a typo here

deadline and see if it comes up with it

press generate and it still worked for

us so that's pretty cool now some of you

have really really complex formula so

it'll be interesting to see how it holds

up but I found that I have pretty good

success by typing that in and getting

the result that I'm looking for now the

last AI feature here that I think is

also pretty powerful is being able to

recommend the appropriate linked record

so I'm inside of a project management

solution looking at some tasks if I

click on this assigning field this is

not assigned to it's not using the users

inside of air table this is actually

pulling from a separate table of people

or of contacts if I edit this field

there's this option to use AI to show

the top matches When selecting a record

so this is not something that actually

happens in the background it's as you're

trying to assign this and it opens it up

it's going to pop up with the

recommended choices and so it allows us

to choose the fields that are important

for it to take a look at both fields

from the task which is kind of the core

record and then the assign it's looking

to the people record so on the task I

wanted to look at the region as well as

the role assigned to that task and on

the people table we wanted to look at

the role that they're assigned as well

as the region quick note that you can

only use up to three Fields here now I'm

kind of curious if you concatenated some

of that text together how it would do

but it might mess with your training

data so I think a little bit more

experimentation needs to be done there

so on my people table I've got my role

or the function in the company that

they're performing and then we also have

the region that they're located in so on

my tasks I created a new task and we've

got it assigned to this Quantum Leap

which is based in latam and we're saying

that this is a task that needs to be

completed by product so if I click on

the field here and open this this is

where we see this top Top matches and

it's sorting it in order of who it

thinks are the top matches that are

available for this now at first I was a

little bit surprised because remember

this is for product and Lam so why does

it come up with Omar who yes he's part

of product but he's in APAC so I think

this really comes down to your training

data what I was kind of hoping it would

do would be to recognize that we have an

association between the regions and

roles and just be able to identify that

we have someone who in latam here let me

go back over to my people and so out of

the product people we've got Arya who is

in latam here so I think she should be

the top of the list of who it recommends

however because she had not actually

been assigned a task yet and so wasn't

part of the training data it didn't

actually recognize that she was the best

fit for it now I've added a record and

it's Lam and it's assigned to Arya so

we've got this Association and if if I

open up the record who does it recommend

no it's someone in AIA so I don't want

to be overly critical here because of

course if we had a large amount of data

this might be able to better detect what

we're looking for but I just say this to

bring up some of the risks of overly

relying on AI is that you need more data

in order for it to make certain

assumptions it's not able to say oh hey

we're looking at these attributes and so

we can infer what it should be based on

the attributes it's because you've got a

volume of data going through the back

end of the system I'd be curious those

of you who have been using this

extensively already if you found that it

has pretty good accuracy for your use

case now at this point you're probably

wondering how do I get my hands on air

table AI there's a limited time free

offering and so you can get these 500

credits to be able to test it out but my

assumption is by using that language of

limited time that this will go away that

it's essentially a free trial because

they're really trying to push everybody

to this paid plan now this is available

if you're on team business or Enterprise

it's going to be $6 per seat per month

if you go to the annual option or else

$7 a month this is all credit-based so

you get 3500 monthly AI credits per seat

now it's important to understand that if

you're on team or business you have to

buy it for everybody in your workspace

you can't just say hey I'd really like

to make it available to these five users

and these 10 users don't need access now

I do want to mention that a lot of the

function ity that we've seen such as

what we're doing in automations could be

accomplished by making a direct call

with open AI API and you could send

information you could retrieve

information you could make updates

yourself you could do things with

generating text now it's really

convenient to have it all within the

application to have it supported by air

table and at six seven bucks a month

it's a pretty good deal but for those of

you who are like I'm already paying so

much in air table licensing there are

other options if you want to use this

feature sparingly now they do give some

suggestions here this would be enough to

do unlimited formulas but how often are

you going to be creating formulas like

it's probably a couple people who are

doing the building and they need access

to that unlimited suggested matching

records again that was helpful because

you're probably doing that one in bulk

70 blog posts or 350 translations so it

gives you an idea of how many credits

this is going to cost but it's still a

little bit hard to wrap your mind around

now I'm on the business plan so I have

this admin panel this is going to look a

little bit different for you if if

you're on the team plan but here if

you're on business I've got Integrations

and development which is under settings

and so here's where you actually enable

the AI features and it's also where you

can view the AI credit usage you can

expand this and it tells you how many

credits that you've used of that Supply

so I'm just testing this in this

workspace I've just been working on this

video this evening and showing a couple

of examples and I've burned through 24%

of that of course normally it would be

on the paid plan and 30 500 credits so

you can start to get a little flavor of

like okay well we saw these examples and

it's burning through that many now if

you do think you're going to burn

through those credits there are options

to be able to purchase additional

credits and so these credits are done in

bulk you can see the pricing with each

of those tiers I'm curious to hear which

AI features are most exciting to you and

how it's going to assist you in your

workflow if you have any questions about

your own air table setup don't hesitate

to reach out to our website at

automation helpers.com where we're

offering free 30 minute

[Music]

consultations