Get a firsthand look at how TeamSupport’s Director of Global Support used AI Agents during a real-world incident.
The AI setup that let my team ignore an incident and just keep helping customers.
A little while ago I posted on LinkedIn about an outage that, for once, didn’t wreck our day. My team kept working their normal queue while the incident got handled around them. No scramble, no all-hands fire drill. The reaction told me I’m not the only one who’s lived the other version of that story. So I want to pull the curtain back, because the how is the whole point.
I’ve been in support long enough to remember when an incident meant everything else stopped. You grab your best people, throw them at the flood, and accept that every other customer is going to wait. That trade never sat right with me, and getting rid of it is something I’ve been working toward for a long time.
Here’s how we got there.
Everything converges in TeamSupport
None of the rest works without this part.
Everything comes together in TeamSupport. Tickets, Salesforce, our knowledge base, Slack, Jira. It all connects here, and it’s from here that we run TeamSupport’s suite of AI agents.
Each one has a job. And each one gets the skills and the direction it needs to actually do that job. Not a vague “be helpful,” but a real role, with real responsibilities, like anyone else on my team.
I train my agents like I train my reps
This is the mental model that makes the whole thing click for me, and it’s my favorite part to talk about.
I train the agents the same way I train a new rep. These are our ticket types. These are our severity levels. These are our groups. This is how and when you escalate. You can literally upload your SOPs and point the agent’s prompt at them, the same documents my humans learn from, and it follows them.
So the agents see what the team sees. They act on it through the lens of their prompt, using the skills I’ve given them, working right alongside our people and our automations.
What that looked like when it hit
When the reports started pouring in, here’s who was doing what.
The Chat Agent was on customer chat, handling conversations and spinning up tickets from them, same as any other day.
The Triage Agent was triaging every ticket landing across all our channels (email, our customer hub, phones) exactly like my human agents do, right alongside them. It’s trained to spot similar reports and build parent/child relationships between them. (It does this for bugs and feature requests too, which gives us far better roll-up visibility into customer and ARR impact, but that’s a post for another day.)
The parent incident ticket itself actually got created straight from our TeamSupport Slack integration. The Triage Agent saw that parent ticket, understood what was related to it, and linked everything up, setting ticket type, product, status, the works, following its training as new tickets streamed in. No manual sorting. Nobody sitting there deciding which of the incoming reports were duplicates.
And then there’s Kevin. Kevin is our Channel Agent, and he was my copilot through the entire thing.
What makes Kevin different is that he can see both sides: what’s happening internally in our team conversations in Slack, and what’s happening in customer conversations inside TeamSupport. So through the whole thing, I just asked him whatever I needed while I ran the response. When I wanted to know which customers were upset and what they were saying, I had Kevin go brief the CSMs, and he pinged their channel himself. And because he’s got skills to act inside TeamSupport, he doesn’t only answer, he does: once an incident wraps, I can have him draft the RCA for the parent ticket from everything he’s gathered.
He’s also just genuinely smart. The base model is strong enough that I can pull near-devops-level intel out of my own conversations. Kevin helps me ask sharper questions and move the needle faster than I would on my own.
The queue ran itself. I ran the incident.
That’s really the whole thing.
The queue ran itself. I ran the incident. And my team just kept helping customers like it was a completely normal day, because for them, it was.
I’ve spent a big chunk of my career trying to give my people their focus back. Letting the machines take the frantic button-clicking so my humans stay in the work that actually matters isn’t a someday vision anymore. It’s just how we work now.
