Artificial Intelligence
Metal Toad University: Class 3
For the rest in the Metal Toad University Series, Click Here.
Metal Toad's team upgraded ToadBot using AWS Bedrock AgentCore, adding multi-person shoutouts, thread detection, and emoji kudos for smarter recognition.
Last week, we wrapped up Metal Toad’s Hackathon 2025, which had AI Modernization as this quarter’s theme. Our team - Anne, David, Nathan, Ingrid, and Calliandra - took it as an opportunity to reimagine one of Metal Toad’s internal tools: ToadBot.
ToadBot has been part of our company culture for years, helping Toads celebrate each other’s work through shoutouts and kudos. This time, our goal was to migrate the tool to AWS Bedrock AgentCore and make shoutouts across Metal Toad smarter, fairer, and more fun.
The problem we set out to solve
We noticed a few areas where ToadBot could improve and make recognition smoother for everyone:
- Limited interaction: Once a shoutout was posted, people could comment, but since ToadBot reads every message in the channel, he would respond with “I didn’t understand” instead of realizing that the comment wasn’t directed at him. This made it difficult to have a natural conversation or celebration in the thread.
- Single shoutout limitation: If multiple shoutouts were posted close together, only the last one counted. That meant only one shoutout could happen at a time, and only one person could be recognized in each message. Shouting out a team or multiple contributors wasn’t possible.
- Performance and reliability: Sometimes ToadBot failed to register points or process shoutouts correctly, and frequent cold starts made responses noticeably slow.
So, we challenged ourselves to fix those issues and explore how AI tools could help make shoutouts smarter, fairer, and funnier.
What we built
We upgraded ToadBot with new AI-powered features designed to make recognition more dynamic and inclusive:
- Multi-person shoutouts: Detects multiple @mentions in a single message and awards points to everyone automatically.
- Thread detection: Allows people to interact directly in shoutout threads on Slack, recognizing replies and avoiding confusion or double counting.
- Toad of the Month: Automatically tallies all points per person and makes it easy for the HR to track weekly and monthly reports.
- Emoji kudos: Introduces special reactions and fun Easter Eggs that make interactions more engaging (and a little funnier!).
Also, the main channel is now cleaner and more organized, old shoutouts stay visible longer, and people can actually get more kudos since recognition doesn’t get lost in the scroll.
|
Before: Shoutouts happened directly in the main Slack channel, and if someone wanted to give points to a colleague, they had to comment outside the thread, which often made the channel noisy and hard to follow. |
Now: All you need to do is react with the MetalToad emoji, and it automatically creates a dedicated thread for that shoutout. Everything related to that kudos (reactions, replies, and follow-ups) stays organized there. |
Technical challenges
Working with AWS Bedrock AgentCore pushed us out of our comfort zone. It’s such a new service that none of the AI tools we used fully “knew” it yet — which meant lots of experimentation, testing, and creative problem-solving.
We had to:
- Teach AI what AgentCore actually is by feeding it AWS Labs examples.
- Tackle Slack OAuth integration, something that sounded simple in theory but turned out to be one of the toughest parts.
Manage performance bottlenecks, since initial deploys took minutes to start and responses could take up to two minutes. Disabling OpenTelemetry and setting up a local testing workaround (thanks, David!) helped us iterate much faster.
Each of these obstacles became an opportunity to learn not only about the technology but also about how AI development often means building the road as you walk on it.
The vibe coding experience
Two of us - Calliandra and Ingrid - joined the Hackathon as non-technical participants, and the team made space for real hands-on learning. We introduced “vibe coding” sessions, where collaboration, and creativity came together in a low-pressure environment. By the end, everyone was experimenting with code, and one of the Easter Eggs in the new ToadBot was even built by Ingrid! It was a great reminder that curiosity and teamwork matter just as much as technical skill when it comes to innovation.
The impact
By the end of the Hackathon, we had a more reliable, engaging, and fair ToadBot that automates recognition and HR insights, reduces manual errors and missed shoutouts and brings more fun, community, and a touch of AI into everyday Slack life.
We’re incredibly proud of what our team built, and grateful for the creativity, humor, and collaboration that defined the experience.
A huge thank you to everyone at @Metal Toad for organizing such a motivating and supportive environment. Hackathon 2025 truly showcased what happens when curiosity, teamwork, and AI innovation come together.
To see more of the hackathon projects, click here.