⭐️ Results:
- Webflow went from rarely responding to achieving a 100% response rate
- We reply to all posts in under an hour
- We now get daily social media reports with key topics.
- I loved Gumloop so much—I joined the company.
The challenge
Before joining Webflow as a product educator, I was an active member of its vibrant community. Once I joined, I realized the sheer scale of Webflow’s presence: over 500K followers across Twitter, LinkedIn, Reddit, and its forum. Big in numbers, even bigger in activity.
This highly engaged community generated over 500 daily mentions across different social channels. Most posts were positive—users sharing templates, asking for recommendations, or celebrating wins. But scattered among them, like needles in a haystack, were users who needed a response from us. A customer caught in a billing issue that could impact their client. A power user running into a product bug.
With a team of two, it was near impossible to identify the posts and the corresponding users where we needed to intervene.
We knew there were community members we needed to respond to—it just wasn’t feasible to find them without draining all our time. We needed a way to separate the posts that required our attention from the ones that didn’t.
Emily Lonetto, Sr. Director of Marketing, Community & Agencies – Webflow
With Gumloop, we ran every post from Reddit, Webflow’s forum, X (Twitter), and YouTube comments through AI to identify which ones required human intervention. Instead of searching for needles in the haystack, the posts that needed us were surfaced automatically.
The results were game-changing. We went from responding ad hoc to confidently ensuring that every user who needed our help received a response—usually in under an hour. By logging data about each post, we could also generate automatic reports on community sentiment and trends.
This workflow was a win-win: the community heard from us when they needed support, and my team no longer lost time wading through endless notifications. The additional data on community trends? The cherry on top.
Emily Lonetto, Sr. Director of Marketing, Community & Agencies
Webflow
Now, let’s dig into how we made it happen.
The solution
We needed a way to analyze each social media post Webflow was tagged in to gauge its sentiment and priority—all without requiring manual intervention.
Starting with Reddit
Reddit was one of our most active—and least moderated—channels, so it was a natural starting point to experiment with Gumloop.
Using the native Reddit scraper node, we automatically pulled in the most recent posts every hour.
With the reddit posts in hand, we used the AI categorizer node to:
- Assess a post’s sentiment (positive, neutral, negative),
- Gauge its priority (low, medium, high)
- Assign it to a category (product feedback, complaint, etc.) with specific instructions that are unique to our product and community.
For us to judge the quality of this workflow, we sent high priority posts directly to Slack for the team to respond right away.
To see if were on the right path, every post’s information (author, content, url, sentiment, priority) was sent to Airtable so we could manually go over the data and ensure that the AI was indeed tagging as we would.
We let this workflow run for a few days to assess whether the AI could be trusted to flag only high priority posts.
The results blew us away: the AI was surprisingly accurate, it only flagged the posts that required our attention. We confirmed this by looking back on all posts logged on Airtable.
This meant that moderating Reddit went from an active task–having someone on the team on schedule to moderate the channel–to a reactive task–jumping in only when necessary. With one less channel to moderate, the team could focus elsewhere on more crucial tasks without letting the community or team down.
Scaling to all channels
The next step was obvious: how do we expand this to all channels?
Since X was Webflow’s most active channel, it was our next focus. At the time, Gumloop didn’t have a native X integration (it does now!), so I built one using the custom node feature and a few prompts to generate the necessary code.
With multiple inputs (X, Reddit), it was crucial to apply the same AI treatment to ensure consistent tagging regardless of the source.
To achieve this, we moved our AI step into a Subflow ensuring that as we scale to multiple sources, we’ll have one place where our posts get tagged and categorized.
We were hesitant that the workflow would scale to X since the posts provide less context. After a few days of testing, to our surprise, the workflow held up remarkably well.
Just like in the case of Reddit, the AI identified only the posts we needed to act on, significantly increasing the amount of posts we responded to while reducing the need to actively monitor the channel.
Outcome
The time savings and impact on the ability to identify the right posts from Reddit and X were so significant that we quickly added posts from the internal forum and Youtube comments.
Before implementing this workflow, we knew we were dropping the ball—not because we didn’t want to engage, but because manually sorting through 500+ mentions per day was impossible.
Now? This AI-powered workflow processes over 500 posts a day and surfaces only the 10–15 that require immediate action.
Instead of drowning in notifications, we know with certainty that we’re responding to every user who needs help quickly and efficiently.
If you’re looking to transform your operations—whether in social media, support, or any other area where automation and AI-driven insights can make a difference—consider following Webflow’s lead. Gumloop allows non-technical employees to achieve the impact of an engineer.
A good starting point to build a social media moderation automation like Webflow is this template flow. This isn't exactly what Webflow has running (their is super customized and tweaked to how they like it) but it's a great initial version to build upon!
Joining Gumloop
When I first started using Webflow in 2019, it gave me a new superpower: the ability to build a fully functional site—just like a developer—without writing any code.
After building this workflow, I had the exact same feeling.
If I could imagine an AI-powered automation, I could simply build it.
It only felt natural to join the Gumloop team and help others unlock this same superpower.
And now, writing this customer story from the other side is an honor!
Learn How to Build This
If you want to learn more about building with Gumloop, I host a weekly learning cohort for free! We teach you the fundamentals of how to automate with AI and even give you free credits for following along.
You can apply to join the cohort here.