Live conversations
Every active conversation across every channel, with intent, status, lock holder, and channel of origin visible at a glance.
Watch routing, agents, channels, and the customer line in real time. Reclaim, reassign, or escalate without forcing your team into the dashboard.
Even when agents reply from their preferred endpoint, owners need a single place to see what's happening across the customer line. The Podium is that place — read-many, multi-supervisor, real-time, and audit-logged on every action.
Every active conversation across every channel, with intent, status, lock holder, and channel of origin visible at a glance.
Who is online, who is paged, who is replying. Set agents to off-hours mode in one click; rotation respected by Matriarch automatically.
See why each conversation went where it went. Sonar inputs, confidence score, policy rule that fired, the chosen route. Click any conversation to expand the reasoning.
Provider quotas (LINE, Meta), webhook delivery success rates, template approval state, error rates per channel. Yellow before red — fix it before it becomes an incident.
From a real demo session, anonymised. The pattern repeats: AI doing the volume, humans handling the ones that need a person, supervisor watching for outliers.
Live conversations: 47 active. 38 in AI mode (AI handling, no agent paged). 7 in HUMAN mode (agent owns, AI watching). 2 in TRANSFERRING (handoff in progress).
Agent availability: Riya (PWA, online), Suresh (WhatsApp, online), Priya (Telegram, away — 4 min idle), Anya (PWA, offline since 14:00).
Channel health: ✓ Web widget · ✓ WhatsApp Cloud API (94% template quality) · ⚠ LINE OA (87% of monthly quota used, 6 days remaining) · ✓ Telegram · ✓ Shopify webhooks.
Recent routing decisions: #C-48201 → Riya (refund forced handoff, confidence 0.91 not relevant) · #C-48198 → AI (pre-sale sizing, confidence 0.96, knowledge cited) · #C-48195 → Suresh (high-cart-value VIP routing) · #C-48190 → Priya (Hindi language preference + complaint policy).
Knowledge gaps surfaced: 3 unresolved questions about return-to-store policy in the last 24h — admin can approve a new FAQ entry from this card directly.
The supervisor reclaims #C-48205 (a customer who explicitly asked for a manager three messages ago that AI hadn't escalated yet — a tuning opportunity caught in real time).
Every supervisor action is logged with actor, timestamp, conversation ID, and reason. Wake supports replay for training, dispute resolution, and compliance.
Step through the routing, the AI's reasoning, the handoff, the agent's reply, and any supervisor intervention. Useful for training, post-mortems, and dispute resolution.
See every action a specific user (supervisor, agent, or AI) took on a specific day, with full context. Useful for performance review and audit.
See every conversation a specific routing rule fired on. Useful for tuning a policy that's escalating too aggressively (or not aggressively enough).
Wake exports include conversation, routing, tool calls, and supervisor actions in a tamper-evident format. CSV, JSON, and direct-to-warehouse options.
The Podium connects through a live event stream to the tenant-scoped state service that owns conversation state. Every state change (new conversation, intent classified, agent paged, lock acquired, conversation closed) emits a delta. The Podium applies the delta locally — no polling, no full reload.
If the WebSocket drops (network blip, mobile sleep, proxy timeout), the Podium falls back to SSE automatically and resumes from the last known event ID. The reconnect is invisible to the supervisor.
Multiple supervisors watch the same tenant simultaneously without contention. Reads are unconstrained; writes (reclaim, reassign, escalate) acquire a per-conversation lock at the routing layer to prevent collisions.
It's a supervisor view, not an agent inbox. Owners and team leads watch the live state of conversations, intent classification, agent availability, and routing decisions across every channel. Agents reply from their own Den; the Podium watches them all.
Yes. Supervisors can reclaim a conversation, reassign it to another agent, push it back to AI, escalate to a different endpoint, or join as a second agent in conference mode for training.
Yes. Every routing decision shows the input signals (intent, sentiment, urgency, customer tags), confidence score, the policy rule that fired, and the chosen route. Useful for tuning policy, training new staff, and explaining a decision to a frustrated customer.
Yes. The Podium is read-many for any user with the supervisor role; writes (reclaim, reassign, escalate) are tracked in Wake with actor identity, timestamp, and reason.
The Podium surfaces real-time signals — what's happening right now. Aggregate analytics (response time, resolution rate, AI vs human split, channel mix, knowledge gaps, common refusals) are available in the analytics view and exportable to CSV or your data warehouse.
Yes — the Podium is a PWA, installable on mobile, with offline transcript caching and push notifications. Useful when the owner is on the road and wants a quick glance.
Yes. Roles are scoped per tenant: owner (full), supervisor (live view + take-over), team lead (scoped to a department), agent (their own conversations only), read-only (audit access without controls). Custom roles supported on Agency.
Demo with your channel mix, agent count, and a real customer message. We'll show how supervisors keep an eye on the line without becoming a bottleneck — and how Wake captures every action for audit.