Using Airsearch

In this video, we walk through how to use Airsearch, Airscale's built-in AI with real-time web access, to automatically extract specific information from company websites. In this example: identifying which companies have a partnership page and retrieving the URL.

0. Why use Airsearch?

Unlike other large language models available on the platform, AirSearch can scrape the internet in real time. This makes it especially powerful when you need to retrieve any information that requires live web access.

In this use case, we imagine being a company looking to partner with others. The goal is to check, for each company in our list, whether they have a partnership page on their website, and if so, get the URL.

1. Start from a company table

The workflow begins with a table of companies, previously built using the Find Companies module in Airscale (tutorials available on the AirSchool page).

Before launching the AI search, make sure your table already contains the columns that will be used as inputs in your prompt. In this example:

• Company name

• Company domain

If an input is referenced in the prompt but does not exist as a column in the table, the workflow won't run.

2. Launch Airsearch

To start, click "Use AI", then:

• Select the rows you want to enrich

• Grant web access (this automatically sets Airsearch as the default language model)

• Write your prompt, or import a saved template

A good Airsearch prompt always includes three elements:

• Connected inputs: each input variable (e.g. company name) must be manually linked to its corresponding column. Click on the variable in your prompt and select the matching column: it turns blue once correctly connected.

• Context: give the AI a clear role and step-by-step instructions. In this example, the prompt asks the AI to act as a B2B market research analyst and explains precisely how to determine whether a company has a partnership page.

• Output indication: specify the exact format expected. Here, the AI is instructed to return Yes if the company has a partner page, or No if not, making it easy to filter results later.

3. Define structured outputs

The "Define outputs" section gives you control over what the AI returns and lets you extract multiple data points in a single run.

By default, the AI returns the main response (Yes / No). To capture additional information:

• Click "Add a field"

• Name the field (e.g. Partner URL)

• Briefly describe what to return (e.g. "Write down the partnership page URL if the company has one")

This way, the AI returns both the qualification answer and the actual URL of the partner page in the same workflow.

4. Review and add results as columns

Once the workflow has run, two new AI columns appear at the end of the table:

• One with "Search completed ✅", & "Click to expand" which reveals the AI reasoning and details of the search
  Note : Any of these fields can be promoted to a permanent column by clicking Add as a column. This makes the data easy to copy into a CRM, share with the team, or use directly in your outreach workflow.

• One with the defined output field, in this case: the Partner URL

By combining real-time web scraping with structured outputs, AirSearch turns a simple list of companies into a fully qualified, enriched dataset, ready to power partnerships, sales, or research workflows.

You can also connect your API key with Airsearch

Prerequisite : you will need to be on a paid plan in order to use Airscale's API

Use the Airscale API endpoint to fetch anything about a company or a prospect from a prompt with some advanced web browsing capabilities.

Find the full tutorial in: Airsearch (AI research agent)