Giles can receive escalations from Archer and Donna, but there's no escalation dashboard or human-view for David/Molly to see and triage escalations in real time. When Archer can't answer, or a customer complaint escalates, where does it surface for humans?
The SHINE Dashboard covers agent logs beautifully, but there's no visual pipeline view showing: leads by stage (new → contacted → quoted → converted), conversion funnel, time-to-contact metrics, or lead value. David and Molly need a "leads cockpit" separate from agent telemetry.
The system generates quotes, schedules calls, and creates orders — but there's no SMS/WhatsApp channel for real-time customer notifications ("Your cleaner is confirmed for Thursday"). Process_Outbound_Queue handles email and Slack, but no mobile channel.
Marshall creates retry tasks for NO_ANSWER and VOICEMAIL outcomes. But there's no smart retry strategy — no max-attempt cap, no escalation after N failures, no time-of-day optimisation. Could lead to infinite retry loops or dead leads clogging the pipeline.
Stripe MCP is connected but not used operationally. No visibility into: payment success/failure rates, subscription churn, revenue per lead source, LTV calculations. Together Property invoicing exposed this gap — commercial invoicing sits outside the system entirely.
Tests are currently manual (fire webhook, check logs). No automated health checks for: agent response times, workflow failure rates, queue processing delays. A "SHINE health monitor" that alerts on anomalies would catch issues before they become customer-facing.
BK Mirror imports visits, but there's no view into provider performance: attendance rates, customer ratings, schedule adherence. Giles handles provider support reactively — a proactive dashboard would flag problems before complaints arrive.
Already in TASKS.md as a line item, but worth highlighting: drip emails are going out without an unsubscribe mechanism. This is a legal requirement (PECR/GDPR) and should be prioritised before scaling email volume.