Last year, I wrote about how we built our CRM in Airtable. Since then, we’ve added a bunch of new AI-driven features. 

Here’s exactly how our CRM enriches data using Airtable’s built-in AI capabilities, all while saving us a ton of time and giving us a quick overview of the person we’ll be talking to (and potentially working with). 

As an Airtable Partner, we believe in leading by example when it comes to using these features in-house, and with the rapid development of Airtable AI features, it’s an easy choice to make. 

And, of course, before we implement systems like this for clients, we always use our own business as a lab to test and iterate to make a perfectly functioning final product. 

New Lead Flow

The start of the process isn’t groundbreaking. Using a Fillout form embedded on our website, the potential client completes their details, and books a Systems Snapshot call. This triggers the Airtable automation. 

1. Contact Data Enrichment in an Airtable CRM

The first point of data enrichment is for the details of the contact. With an AI step within the Airtable automation (using GPT-5), we use a prompt that returns these data points:

1. LinkedIn profile link

2. Summary

3. Profile Image

4. Job Title

Previously, we used an API to achieve the same thing, but when LinkedIn came down hard on these platforms, the API was deprecated, so AI was the next best thing. The other thing is that not all of the information is always accessible on LinkedIn, so this enables us to search other sources for the same data.

2. Company Information

For companies, we use a similar process, except we’re seeking a lot more data points. This is definitely a time saver. These are the data points:

1. Sector

2. Favicon (URL)

3. Logo (URL)

4. LinkedIn (URL)

5. Year founded

6. Company domain

7. Country

8. Suburb

9. State

10. Postcode

11. Company description

Using this information, we’re starting to build a profile on the company, and avoiding a lot of data entry.

3. Lead Motivation Research

The third AI step in the automation is a bit more abstract. We’re not looking for particular data points, we want to know:

1. Buying motivations

2. Important Values

3. Best approach to selling services to them (the lead)

The output from this step is pretty variable and it’s more of a nice-to-have than a key part of the CRM infrastructure. To me, it’s information that serves as a conversation starter, and the real value lies in the conversations with the potential client.

I’ve chosen to put this in its own step, because it takes a bit of time to run and risks timing out if it is combined with other requests. 

Observations So Far

It’s been running quite well since it was implemented, however I do have some observations and ideas about how it could be improved.

• The automation takes a while to complete. However, since it’s not really a time-sensitive part of the system, it doesn’t really matter.

• We use two scripts to parse the output from GPT-5. These could potentially have been combined, but keeping in separate steps avoids timeouts, and is easier to debug if it comes to that.

• The system could do with some error handlers, in case of unexpected AI responses, or script errors.

• We have the luxury of a lot of AI credits, so we can use a more expensive model. One of the lower priced models would likely work fine. Maybe there is a better model completely, I haven’t experimented with it.

Overall, I’m really happy with it. I'm excited to see what we can do next with AI within Airtable.