Its internal AI agents now save employees more than 1,200 hours per month across sales, support, product and engineering. The systems answer over 5,000 questions monthly and also perform tasks such as assembling customer demos and opening support tickets.
The project began with Slack assistants that retrieved retention figures, documentation and account history. It expanded into specialized agents: a go-to-market bot answers customer questions, a forward-deployed engineering bot handles product questions and builds configured demos, and a triage bot turns bug reports or feature requests into Linear tickets.
The triage workflow created more than 150 tickets in the latest month. Before opening one, it searches Slack for context, checks for duplicates, looks for enough diagnostic detail and uses an ownership document to route the issue. Data-analysis agents answer questions inside domain-specific channels, where staff can inspect the response and ask follow-ups.
These are company-reported productivity figures rather than an independent time study. ElevenLabs also says reliability depends on carefully limiting what each agent can see. Its largest retrieval-quality gains came from removing unnecessary schema fields, not from more prompt engineering. The lesson is practical: narrower, well-maintained data access can matter more than a more elaborate instruction block.