“I was pretty blown away. I think I saved myself a day or two of effort. Now it’s running on autopilot.”
— Justin Gallagher, Co-Founder of Hoop
As a startup’s user base grows, so does your support volume—and with it, the risk of missing critical requests or losing touch with customer needs.
That was the exact challenge we faced at Hoop. We needed a bird’s-eye view of reported issues, feature requests, and overall product feedback.
Enter Justin Gallagher, my fellow co-founder, who decided to tackle this opportunity with an AI first mindset. Yes, we’re an AI company using AI to build our AI product. Meta, right?
In just a few hours, Justin hacked together a daily Slack summary bot to keep our entire team in the loop about what we were hearing from customers in support, saving us from hours of manual summarizing and/or looking for a dedicated tool to solve this problem.
The problem: too many chats, too little insight
We manage our customer support through Intercom, typically via real-time chat on our website. It’s a fantastic channel because it's super fast and convenient. However, as we grew, the number of conversations skyrocketed.
- Manual approach: We’d try to periodically go in, read everything, piece together common issues, only to realize we missed half the context.
- Existing tools: Intercom does have analytics, but we wanted our data synced to Notion, plus daily highlights in Slack—fine-grained control we couldn’t quite get “out of the box.”
“We were losing touch with all the cases… I wanted to make sure that didn’t happen,” Justin explains.
The first steps: Postman, API calls, and AI summaries
Justin’s initial plan was a one-time summary of that week’s support cases. He discovered:
- Intercom API: A quick search revealed Intercom does indeed have an API.
- Postman: A handy (free) tool for making API requests and exploring data. Postman’s built-in AI feature transformed JSON data into table format, making it far easier to read.
Now with all the data in a table, Justin copied it into ChatGPT and asked:
“Here are our support tickets. What are the top trends this week?”
He tweaked his prompt to ask for more structured results (e.g., referencing specific ticket IDs). Within minutes, he had a quick summary, which he dropped into our weekly meeting agenda. Cue the “Aha!” moment.
Going further: daily slack summaries on autopilot
A single summary was nice, but we needed a daily view without manually repeating the entire process. Why not automate? Justin used ChatGPT to write code that integrated:
- Intercom’s API to fetch new support tickets
- OpenAI’s API to summarize the data
- Notion’s API to sync known features/bugs
- Slack’s API to post a daily summary in a channel
- Iteration & debugging: ChatGPT handled the majority of the coding, though Justin had to confirm some details or fix quirks. For instance, Intercom’s pagination can be finicky (only 150 records at a time). ChatGPT wrote an initial approach incorrectly, but corrected itself once Justin pointed out the official doc requirements.
- Deploy on Replit: After the script worked locally, Justin asked ChatGPT: “How do I schedule this to run daily without installing anything on my machine or paying money?” ChatGPT suggested Replit. Justin uploaded his script there and set up a daily schedule.
Now, every evening, the bot automatically:
- Grabs the day’s support tickets from Intercom
- Cross-references known issues from Notion
- Feeds relevant data + a prompt into OpenAI
- Posts a tidy summary to our support Slack channel
“We’ve got a running feed of the top support trends every day, no manual lifts.”
Lessons learned (and surviving AI’s quirks)
Despite the success, Justin hit a few rough edges:
- Hallucinations & mistakes: At times, ChatGPT would guess parameters incorrectly or revert previously fixed code. Catching those errors required some human debugging.
- Token/output limits: Longer scripts (250+ lines) often got truncated by ChatGPT. Justin had to prompt it cleverly or piece the code together manually.
- Manual “checkpoints”: At each step, he tested the script, verifying the logic—especially around pagination. “It’s not fully hands-off,” he admits, “but it’s a huge time-saver.”
Why this matters: building the future, faster
We often talk about AI as a superpower for repetitive tasks like summarizing. But the real magic happens when AI helps us build the glue between our apps. Justin’s daily Slack bot is a prime example: a few hours, some iterative prompts, and voilà—a consistent pulse on customer needs, automatically. "As the models get more advanced, it’ll be insane what you can do in minutes,” says Justin.
Key takeaways:
- You don’t need to be an expert coder, but some comfort with APIs is helpful.
- AI can do the heavy lifting for grunt code tasks—though a human eye (or two) is still crucial.
- Tools like Replit and Postman are your best friends. They’re free, user-friendly, and pair well with AI’s code generation.
- Start small. Summaries, daily logs, or simple “trends” are great first automations.
What’s next?
At Hoop, we’ll keep refining how we integrate AI into our workflows and we promise to share more behind-the-scenes experiments soon. Like so many of you, we're learning every day how to make AI tools more useful in our lives. So go ahead: Grab your coffee, open up ChatGPT, and see how it can cut hours of manual drudgery out of your day.
