Flow basics
Welcome to Gumloop 101! In this lesson, you'll learn the steps to automating your first workflow in Gumloop. We'll use a real world example but no matter what you're automating, the building blocks are the same.
For this whole course, imagine you’re a conference organizer wanting to recommend the perfect talks to every attendees.
You’ve got a list of attendees, a packed agenda of talks, and one big goal: sending each attendee a personalized email recommending the sessions most relevant to their professional background. Sounds like a lot of manual work, right?
That’s where automation—and Gumloop—comes in.
Defining your workflow
Before diving into building anything, it’s important to take a step back and think through the workflow. In Gumloop (and really, in any automation), everything comes down to three key stages: data in, data transform, data out.
- Data in: We’ve got an attendee’s email address and a list of available talks at the conference.
- Data transform: This is the magic part. First, we enrich the attendee’s profile—pulling their work history from LinkedIn. Then we ask an AI model to recommend the three most relevant talks for them based on that data.
- Data out: Finally, we use that recommendation to generate a personalized email.
In a diagram (and in Gumloop!), this workflow looks like this:
Let’s build it in Gumloop
In Gumloop, we call these automations flows. Each flow is made up of nodes, which are the individual steps your automation takes.
So where do we start? Right at the beginning—with the Input node. This is where we capture the attendee’s email.
Next, we enrich that email using the Enrich Contact and the LinkedIn Profile Scraper nodes. Each node in Gumloop follows the same basic structure:
- Inputs: This is the data coming into the node (in this case, from the Input node).
- Parameters: These are the options you configure—like selecting what info you want back from the enrichment step.
- Outputs: The result of that node, which you can use downstream.
Together, the Enrich Contact and Linked Profile Scraper nodes provide us with the attendee’s work experiences.
Asking AI for Recommendations
Gumloop offers many ways to use AI in different AI nodes.
The Ask AI node works similar to your favorite AI client (ChatGPT, Claude) where you prompt a language model and it returns a response. We want it to recommend three talks based on both the attendee’s work experience and the list of conference talks.
You can pass in information from earlier nodes—like the LinkedIn data—so the AI has everything it needs to recommend the right talk. In Gumloop, you can use any model you like (we’ve got partnerships in place with all the major providers).
Drafting the Email
Now that we’ve got our recommended talks (from the Ask AI node), the recipient (from the input node), let’s turn them into an actual email. That’s where the Gmail Sender node comes in.
For now, we’ll keep it in draft mode—just in case we want to double-check or tweak anything before hitting send.
And just like that, you’ve automated a fully personalized experience for each attendee.
Now if we're being honest, this workflow isn’t quite automated since it only works for one attendee at a time! In the next lesson, we’ll learn how to run this on all attendees so we can draft hundreds (thousands!) of emails at once for all attendees.
