From Repeated Questions to Instant Answers

A mid-sized financial services company had a familiar problem: the same questions kept getting asked, and the same people kept answering them.

The DevOps team — about 10 engineers — were fielding constant queries from across the business. "How do I set up my dev environment?" "Where's the runbook for X?" "What's the process for Y?" The answers existed somewhere — buried in Slack threads, GitHub repos, internal wikis, and people's heads — but nobody could find them quickly.

The IT support team of 20 had it worse. They were handling thousands of tickets a month, many of which were variations of the same handful of questions. Senior engineers were spending hours on repetitive support instead of doing the work they were hired for.

The bot was built on a reusable framework — standing up a new instance for a different team was straightforward, each with its own Slack channel, GitHub repos, and documentation sources. Within a few months, 16 teams across the business had their own bots. Nobody mandated it. Teams saw it working and asked for their own.

For IT support, about 40% of queries were now handled entirely by the bot, with another 20% partially assisted. That's a significant chunk of repetitive work that no longer lands on someone's desk. Engineers got their time back, and employees got faster answers. Across the IT Support and DevOps teams alone, the bot delivered over $200K AUD in annual efficiency savings — freeing skilled engineers from repetitive questions and reducing support bottlenecks, with further gains across 14 additional teams that adopted the system organically.

The bot kept learning. As teams used it and fed back on answers, accuracy stayed above 97%. The YAML configuration approach meant teams could adjust the bot's knowledge and behaviour themselves — Boxcar built the capability, not the dependency.