About Human Behavior
Human Behavior is a product analytics and testing tool launched this week that moves beyond traditional dashboards. It combines session replay capture with AI agents that watch those replays, identify user struggles, and then act on them directly. The product states that product analytics told you what happened, while Human Behavior handles it by closing the loop between observation and action.
Review
Human Behavior positions itself in a growing space of AI-driven product analytics, where the founder's own dogfooding numbers are the most concrete evidence of what this tool does. The stated workflow has four stages: collect events, errors, and session replays via an SDK; have AI watch every replay to catch rage clicks, dead buttons, and silent give-ups; let background agents take action by emailing the customer, updating Linear and CRM entries, and opening pull requests with replay evidence; and finally loop back so agents check the results. The entire system lives in Slack or SMS, not in a dashboard.
The founder reports the internal use appears to be a notable anecdote: their agents watched 15,000 sessions and opened 16 PRs for issues like dead buttons and layout shifts without any of those surfacing as prior bug reports.
Key Features
- SDK capture of events, errors, and session replays
- AI vision model (Gemini 2.5 Flash) watches session replays and detects user frustration signals
- Background agents that email customers, create Linear issues, and open pull requests on GitHub with replay evidence attached
- Agent-driven results checking that feeds improvements back into the product without human monitoring
- Reporting and updates delivered through Slack, SMS, and WhatsApp rather than a dashboard interface
Pricing and Value
Pricing details are not published on the product page. The company's page lists a free option, but specific tiers, limits, and per-seat costs are not defined in the public reference. Sign-up appears to happen through booking a call via humanbehavior.co, per the founder's launch comments. The value story focuses on the cost of unread session data and unattended bug reports; the founder's own metrics, like 16 PRs from 15,000 watched sessions, serve as the reference point.
Pros
- Agents act on replays rather than just surfacing findings, with Linear updates and PRs generated directly
- Background agents run around the clock without requiring dashboard supervision, as orchestrated by Trigger.dev
- Session replay evidence is attached to every PR and issue, giving developers the context to reproduce bugs quickly
- Interfaces through Slack and SMS, placing alerts where teams already talk
Cons
- Customer-facing emails sent by agents can't be reversed if they're wrong, and the reference does not show whether a human approval step exists for those actions
- Vision model accuracy on nuanced user intent still has limits, so false positives may misclassify an intentional interaction as a rage click or a give-up signing from a user without consent
- The tool is not well suited for organizations needing a human review before automated actions, such as regulated or very small teams without a track record of trusting replay footage; they'll need to set approval guardrails manually
For teams who already rely on Slack and Linear, Human Behavior offers a closed-loop workflow that fits in those tools; for those who prefer dashboards or prefer a human gate, it sits at a distance. The most concrete use case is a product team willing to let an AI coding agent open PRs with replay evidence and then audit the merge log.
Open 'Human Behavior' Website
Your membership also unlocks:
