Skip to content
Ittisal
Platform

منصة واحدة لكل المحادثات

One routing engine behind every conversation

Agent Desktop, unified queues and the GCC AI engine share the same tenant context, the same capacity budget and the same flow graph. Nothing is bolted on after the call has started.

10 channel types130 Studio node typesDeployable in-boundary
Three parts

Agent Desktop, Unified Queues, GCC AI Engine

Each part is useful alone and cheaper together, because routing decisions, transcripts and dispositions never have to be reconciled across systems.

Agent Desktop

A single surface for calls, WhatsApp threads, email and chat. The agent receives live transcript, sentiment and a disposition draft on escalation, so the customer is not asked to repeat anything.

live transcript · disposition draft · CRM panel

Unified queues

One capacity budget per agent, spent by channel: a call takes all 100 points, a WhatsApp thread a quarter, an email ticket a tenth. Skills and language are matched before assignment.

100 capacity points · skill and language match

GCC AI engine

Recognition, generation and synthesis tuned for Gulf, Levantine and Egyptian dialects with code-switching into English. Managed or self-hosted, selected per flow.

dialect aware · managed or in-boundary
Agent surface

What the agent sees mid-call

Voice · active+971 50 •• 4417
100 pts
Live transcript · AR / EN
customer · ar
السلام عليكم، عندي مشكلة في فاتورة هذا الشهر.
Sentiment
46Neutral
Copilot suggestions
listening…
Arabic strings are set in the product's Arabic face and always run RTL · transcript shape follows InteractionTask.transcripts

Interactive preview — step through the escalation the way an agent receives it.

Execution

How a conversation moves through the stack

  1. 01

    Channel arrives

    The trunk or Meta webhook hands the conversation to the media layer, which asks the flow engine what this number or thread should do.

  2. 02

    Flow decides

    The published graph runs node by node. Media and inference planes open only when a node needs them.

  3. 03

    AI turn or queue

    A voice-agent node forks audio into recognition; a queue node matches skill, language and remaining capacity.

  4. 04

    Resolution written back

    Transcript, sentiment, disposition and CRM writes land through the graph, not a side channel.

Operating numbers from production tenants

< 800 ms
voice turn latency, p95
62%
typical AI deflection after tuning
99.99%
contracted platform uptime
4
GCC hosting regions available

See the platform answer one of your own queues

A scoped pilot on a single queue, with your dialect, your CRM and a success metric agreed in advance.