Customer support · Production

Browserbase builds a generalized internal AI agent (bb) that automates knowledge work across engineering, ops, sales, and support

The problem

Browserbase had manual workflows for session investigation and PR writing, with no unified automation layer for knowledge work that happens in web apps without clean APIs.

Workflow diagram · grounded in source
1
User invokes bb in Slack
trigger
“Someone types @bb in a Slack channel. The Slack events handler dispatches to an internal endpoint”
2
Webhook event triggers background run
trigger
“External events like a Pylon support ticket closing or a Circleback meeting transcript landing trigger webhook handlers”
3
Agent self-selects skills
routing
“For interactive Slack sessions, the agent reads this routing table and self-selects”
4
Agent executes task with tools
ai_action
“@bb lives in Slack and can write PRs, investigate production sessions, query Snowflake, log feature requests to HubSpot, and run browser agents across engineering, ops, sales, support, and exec”
5
Results streamed to Slack
output
“The agent streams SSE events back, which get translated into real-time Slack message updates”
Reported outcome

The agent bb runs the feature request pipeline at 100% coverage with zero human effort, reduced first response time by 99% to under 24 hours, and cut session investigation from 30–60 minutes to a single Slack message.

Reported metrics
Feature request pipeline coverage100%
Human effort for feature request pipelinezero human effort
First response time reduction99%
First response time target<24hrs
Show all 6 reported metrics
feature request pipeline coverage100%
human effort for feature request pipelinezero human effort
first response time reduction99%
first response time target<24hrs
session investigation timedropped from 30–60 minutes of manual log-diving to a single Slack message
engineer PR writingMany engineers went from writing a majority of PRs by hand to reviewing them
Reported stack
OpenCodeSlackSnowflakeHubSpotPylonCirclebackGrafanaTinybirdLokiNATSLinearGitHub CLITailscaleBrowserbase Platform
Source
https://www.browserbase.com/blog/internal-agents
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The agent bb runs the feature request pipeline at 100% coverage with zero human effort, reduced first response time by 99% to under 24 hours, and cut session investigation from 30–60 minutes to a single Slack message.

What tools did this team use?

OpenCode, Slack, Snowflake, HubSpot, Pylon, Circleback, Grafana, Tinybird, Loki, NATS.

What results were reported?

Feature request pipeline coverage: 100%; Human effort for feature request pipeline: zero human effort; First response time reduction: 99%; First response time target: <24hrs (source-reported, not independently verified).

How is this customer support AI workflow structured?

User invokes bb in Slack → Webhook event triggers background run → Agent self-selects skills → Agent executes task with tools → Results streamed to Slack.

Related customer support cases