Booking request after hours
- Customer signal
- Customer asks for an appointment slot at 11pm.
- Routing decision
- AI checks calendar availability, proposes the next two slots, holds the reservation, and sends a confirmation. No human paged.
Category definition
What is a semantic PBX?
A semantic PBX routes customer conversations by meaning, not menu choice. Intent, context, knowledge, policy, and team availability decide whether to answer, ask, run a tool, or hand off.
Plain-language definition
Routing modern conversations the way people actually communicate
Phone-era PBX systems route calls through extensions, queues, and IVR menus. Today your customers write to you on six channels at once, attach photos and voice notes, and expect continuity across all of them. A semantic PBX is the routing layer for that world.
Phone PBX
Routes calls. Inputs are extensions, queues, and IVR menus. Optimized for voice and call centers. Adds zero context to a missed call.
Semantic PBX
Routes conversations. Inputs are intent, knowledge, policy, and availability. Optimized for messaging across web, WhatsApp, Telegram, LINE, Meta, Shopify. Captures the full Wake.
Chatbot or shared inbox
Either answers individual messages or just collects them. No routing intelligence, no consistent handoff, no audit-grade transcript across channels.
Architecture
How a semantic PBX is composed
Every conversation flows through five layers. Each can be configured per tenant.
- 01
Channels
Web chat, WhatsApp Cloud API, Instagram, Messenger, Telegram, LINE, Shopify, custom webhooks. The PBX listens on every entrypoint.
- 02
Intent + context
Sonar reads message content, conversation history, customer profile, and channel signals to score intent and confidence.
- 03
Knowledge + policy
The routing layer cross-references your knowledge base, business hours, escalation policy, and live tool availability.
- 04
Decide + route
Answer, ask a clarifying question, run a tool, or hand off to the right human on the endpoint they already use.
- 05
Wake the audit trail
Every routing decision, tool call, and handoff is captured for analytics, replay, training, and compliance.
Example flows
What it does in real conversations
Four common scenarios where a semantic PBX outperforms a chatbot or a shared inbox alone.
Refund or complaint
- Customer signal
- Customer message contains 'refund', 'broken', or 'wrong order'.
- Routing decision
- Policy escalates immediately. The right agent is paged on their preferred channel with the order details, transcript, and a recommended response template.
Product question with low confidence
- Customer signal
- Customer asks a niche technical question the AI is not sure about.
- Routing decision
- AI offers to check with a person, captures contact preference, and routes to the human on a channel they already use.
Returning customer
- Customer signal
- Customer continues a conversation from yesterday on a different channel.
- Routing decision
- Identity is matched. The Wake transcript carries forward. The same agent or pod is paged where possible.
When not to use it
Honest limits
- You only do outbound campaigns. Use a marketing/messaging platform. A semantic PBX is built for inbound conversations.
- You want a full helpdesk with SLA workflows. A semantic PBX can integrate with a helpdesk; it is not a replacement for ticketing-heavy ops.
- You replace human judgement entirely. The point is augmenting humans, not removing them. Risk-sensitive intents always need a person.
- You expect "set and forget". Routing policies, knowledge, and channel constraints need real review as your business evolves.
More on semantic PBX
Frequently asked questions
What is the difference between a semantic PBX and a chatbot?
A chatbot answers individual messages. A semantic PBX manages the whole conversation: intent classification, knowledge lookup, policy enforcement, human handoff, transcript continuity, and audit trail across every channel.
Why not just route by phone number or extension?
Phone-era PBX assumes a caller picks an option from a menu. Modern customers send messages with full context — orders, screenshots, voice notes — across web, WhatsApp, Instagram, and Telegram. Routing by meaning beats routing by menu choice.
How is a semantic PBX different from a shared inbox?
A shared inbox centralizes messages but still requires staff to read, classify, and respond manually. A semantic PBX answers routine intent, escalates valuable conversations, and only forwards what actually needs human judgement.
When should you not use a semantic PBX?
If your team replies manually to every message and prefers a single dashboard workflow, a shared inbox may be enough. If you need bulk outbound campaigns at scale, a marketing platform fits better. A semantic PBX is for inbound conversations that benefit from intelligent triage and routing.
How does OrcaLinq implement semantic PBX?
OrcaLinq runs an AI-first answering layer (Matriarch routing engine) on top of a Sonar semantic intent signal, a knowledge layer (Lore), policy/availability rules (Tides), agent endpoints (Den), and a Wake transcript that captures every routing decision. Channels remain BYOK — your provider, your credentials.
See a semantic PBX run on your own conversations.
Book a 30-minute working demo. Bring a real customer message. We'll show what an AI-first answering layer plus human handoff looks like for your team.