When you open a customer support ticket, it often feels like dropping a message into a void. Someone will eventually respond, maybe the issue gets resolved, and then the record disappears into a database that few ever look at again.
For Tom Bachant, founder and CEO of Unthread, that seemed like wasted potential. Every one of those conversations contained insight about how people actually experience a product. Patterns were buried inside, waiting for someone or something to uncover them.
Today, Tom is showing teams how to do just that. His company helps organizations handle internal IT and HR requests directly inside Slack, transforming an endless stream of messages into structured, searchable support. But his latest experiment, the Support Ticket Analyzer Agent, goes a step further. It uses AI to comb through ticket data and reveal what matters most.
Tom’s story as a builder started long before AI became the buzzword it is today. While studying at the University of Connecticut, he launched a ride-sharing app called Dashride, years before Uber and Lyft dominated the space. The idea was simple: Students needed an easy way to get around campus safely, and he had the curiosity to build something that could help.
That project became a full company, later acquired by Cruise, where Tom stayed on as an engineering manager. He learned what it means to ship products that serve thousands of people at once and how small bits of automation could change an entire operation.
By the time he founded Unthread, the through-line in his work was clear. He loved solving messy, human problems with elegant technology.
“I’ve always been a hacker type,” he told me. “The fun part is seeing how a few lines of code can create something that changes how people work.”
Anyone who works in Slack knows the chaos that can unfold. Important messages get buried. Requests vanish into long threads. The same question gets asked three times a week.
Unthread’s mission is to make that manageable. It acts as an AI-automated help desk built inside Slack. If someone asks for a new laptop or a password reset, Unthread can capture the request, track its progress, and automate the response.
But Tom saw something deeper happening. Each message and each ticket was a fragment of operational truth, evidence of how teams actually functioned. What if that information could be analyzed to reveal bigger patterns?
That question led to the Support Ticket Analyzer Agent, now available on Agent.ai (go try it, it's free). The tool accepts a simple CSV export of support data, no matter whether it comes from Zendesk, Jira, HubSpot, or Unthread itself, and uses large language models to detect recurring themes, sentiment, and urgency.
“Having AI summarize data is easy,” Tom says. “Making that summary actionable is the hard part.”
The agent doesn’t just count how many times an issue occurred. It ranks those issues by impact. It highlights urgent problems, surfaces opportunities for documentation, and even identifies where teams could automate repetitive tasks. For anyone responsible for customer experience, product management, or internal operations, it’s like handing your backlog to a data-literate teammate who already understands context.
Tom first built the tool for Unthread’s own use. His team ran thousands of internal support tickets through it, comparing the AI’s analysis with their own judgments. The results were surprising.
“I realized I didn’t need to babysit the system,” he said. “It understood the same patterns I saw and sometimes communicated them even more clearly than I could.”
Every builder dreams of feedback loops that improve themselves. For Tom, the Support Ticket Analyzer became one. His team now reviews the output in sprint planning sessions to decide which issues to fix or document next. What once took hours now happens in seconds, freeing them to focus on action instead of organization.
But perhaps the most interesting choice came next. Tom decided to share it with everyone. The prompt behind the agent, the logic that tells the AI how to interpret ticket data, is available for anyone to use through Agent.ai. Even if you have never touched Unthread, you can upload your own CSV file and see what insights emerge.
That openness fits a pattern in Tom’s work. He believes good AI tools should not lock people into one ecosystem. They should meet users where they already are.
“You shouldn’t need to change systems to benefit from automation,” he explained. “If AI can understand your data, it should work with whatever tools you already use.”
For users exploring AI for the first time, Tom’s story is an approachable blueprint. He didn’t start with a grand vision of replacing help desks or rewriting workflows. He started with a single, recurring problem: too much support data and too little time to interpret it.
That narrow focus kept the project grounded. Each iteration of the agent solved one more piece of the problem, first summarizing data, then prioritizing issues, and finally suggesting next steps.
It also taught Tom a critical lesson: AI is most effective when paired with human judgment.
“The agent gives you a great starting point,” he said. “But we don’t follow it blindly. It helps teams align faster so they can spend their energy on decisions, not on reading hundreds of tickets.”
In practice, that human-plus-AI approach is what makes the Support Ticket Analyzer resonate with real users. It doesn’t try to replace people; it makes their judgment more scalable.
The next problem Tom wants to tackle is one every company recognizes: outdated documentation.
“We all have those Confluence pages or Google Docs that haven’t been touched in years,” he laughed. “Someone answers the same question in Slack, but the docs never get updated.”
Unthread’s upcoming Documentation Generation Agent aims to fix that. It will analyze existing documents, compare them to real conversations, and automatically suggest updates or create new articles when gaps appear. The goal is not just to save time but to keep institutional knowledge alive.
As Tom described it, the agent will do the tedious part—finding and drafting updates—while humans simply review and approve the changes. It’s the same philosophy that shaped Unthread’s support tools: let AI handle the heavy lifting so people can focus on the work that matters.
For many, the idea of building an agent still feels intimidating. Tom’s story proves that it doesn’t have to be. He didn’t build the Support Ticket Analyzer as a massive platform feature. He built it as a useful experiment. By solving a real problem for his own team, he ended up creating something valuable for others.
That is the magic of agents. They start small, maybe as an automation, a prompt, or a script, and evolve into something that can genuinely improve how people work.
Whether you manage customer support, run operations, or just want to understand your organization’s noise, Tom’s agent shows what is possible when AI turns listening into learning.
If you've ever wondered what your support tickets might be trying to tell you, now is the time to find out. You can try Tom Bachant’s Support Ticket Analyzer Agent directly on Agent.ai. Upload a CSV file from your system, let the agent do the reading, and see what insights surface.
You might discover your next product idea, process improvement, or documentation update hiding in plain sight, inside the conversations you already have every day.