Capturing, updating, and reusing support knowledge with AI

Hey everyone - just shipped TicketReel, a Canvas Kit app I built for the Intercom inbox.

The idea: when an agent resolves a ticket, they add an internal note with the resolution steps, then click Generate Video in the TicketReel panel. The app turns that note into a 60-second branded AI avatar video — scripted by GPT-4o, rendered via HeyGen, captioned via Whisper — and posts the video link back to the conversation as an internal note.

The video can be shared with the customer or published to your knowledge base, so the same resolution helps future customers who hit the same issue.

Built with Canvas Kit + the Conversations API. Happy to answer any questions about how it works under the hood.

App Store listing:
https://ticketreel.chriscastle.com/

Knowledge copilot helps you maintain your knowledge base by identifying content gaps, suggesting updates, and generating draft articles or procedures from ticket data. It provides health metrics like coverage, freshness, and AI readability to assess your content. Use conversational assistance to manage tasks, review recommendations, and create or update articles, keeping your knowledge base current and optimized for AI use.

Yesterday we unveiled Operator, an agent that runs your customer operations. Operator spots problems, finds opportunities to improve, and it acts on them – tuning Fin, keeping your knowledge current, and scaling automation as your business changes. Here's Brian , VP of Product at Fin, explaining why Operator is a paradigm shift for how you interact with business software. You can watch the full Operator launch event at the link in the comments.

Your customers don't want to talk to AI.

When systems are down and the pressure is on, they want to know there's a real person on the other side. Someone who gets it, who can pull in the right people, and who won't tell them to "try asking the chatbot."

I've heard this over and over. And I agree with them.

But here's the thing. The best support teams I've built aren't choosing between humans and AI. They're using AI to make the humans better.

I started in support as one of two people covering 24/7 for a global client base. Every ticket was manual. Every escalation was a phone call. Every piece of tribal knowledge lived in someone's head.

Seven years later, I've led that same function through multiple mergers and acquisitions, platform integrations, and a full operational transformation. I didn't do any of this alone. I had a team that was willing to try new things and push through the growing pains.

We replaced manual NOC monitoring with automated alerting. We used AI-assisted tooling to break down knowledge silos and actually capture what people knew. We gave support engineers their own troubleshooting toolkits so they could resolve issues without waiting on engineering every time.

And we saw it in the results. Resolution times improved quarter over quarter. Support self-solved rates increased, which relieved load on backend engineering resources.

The result wasn't just better metrics. It was a team that could take on harder, more complex work because they weren't buried in the repetitive stuff.

I know the future of support feels scary right now. But the unlock isn't replacing people with AI. It's AI-assisted human agents. People who are hungry to grow, supported by tools that surface the right data at the right time. People who learn to use AI responsibly, who watch for hallucinations, who make sure the output is actually correct.

And the humans? They level up. They build a real foundation for a growing career in our industry.

Let's make it better for future us's.