Campfire brings AI drafting, policy context, and human approvals into one calm operating surface. The page below shows the direction production screens should move toward.
Similar cases resolve well when the AI drafts the exception and a human approves the payout destination change.
Human sign-off required before sending the final refund confirmation because destination details changed.
Confidence, escalation pressure, and policy guardrails stay close to every draft so teams can trust what is happening.
Queue, thread, context, and approvals sit in one composed operating view instead of a stack of generic panels.
Escalation, assignment, and approvals are treated as first-class moments, not buried utility actions.
visible confidence gating
faster without panic UI
used when nuance matters
Fewer generic panels. Stronger sequence. Operational states become visible. Campfire starts to feel like a differentiated support product rather than a polished starter kit.
The rollout is staged so the product can ship visible improvements without waiting for a full redesign. Start with the highest-traffic surfaces, then harden secondary flows.
Lock the new visual thesis, typography stack, page rhythm, surface hierarchy, and motion rules before any broad migration.
Token audit, page grammar, component tiers, screenshot treatment, approval samples
Rebuild homepage, pricing, and features around product theater, proof, and a sharper buyer narrative.
New hero, trust section, product story sequence, pricing narrative, responsive polish
Apply the same language to dashboard, inbox, and conversation detail so the operational product feels authored and premium.
Dashboard shell, inbox master-detail scene, KPI hierarchy, approval and escalation states
Roll the direction across settings and secondary surfaces, then tighten motion, accessibility, density, and testing.
Secondary page normalization, a11y pass, motion review, regression screenshots
The direction is ready to migrate into homepage, pricing, dashboard, and inbox surfaces. Each phase should preserve the same rule: AI is useful only when the human can see, approve, and override it.