Skip to content

The platform

One agent.
Every channel.
Real work done.

Five parts that share a brain: the same knowledge, the same conversation logic and the same integrations — whether the customer called, messaged or emailed. It answers, and then it does the thing behind the answer.

A live conversation — watch it resolve, in Arabic and English
WhatsAppHandled by AI
Try

Why it's different

A chatbot answers. Heard does the work.

Most AI support is a search box with better manners. The value was never in answering the question — it is in doing the thing behind it, correctly, without a human, and then proving it did.

  • One agent across voice, WhatsApp, web, email and SMS — one brain and one knowledge base, not five disconnected bots that each know half the story.
  • It takes real actions inside the conversation — opens a claim, checks an order, books an appointment, issues a refund — not just replies with what you could do.
  • Every answer is grounded in your own knowledge and cited, with multi-stage validation before it speaks, so it does not invent policy or price.
  • It hands off to a human with the full conversation the instant it should — and every correction makes the next answer better.
  • Native Arabic, including Gulf dialect, and English — the same agent, switching to whatever the customer speaks.
  • Live in under a day: upload your knowledge, connect a number or channel, and it starts resolving real conversations.

The architecture

Five surfaces. One brain.

The channels are not five products. They are five ways to reach one intelligence — which is why the customer never has to repeat themselves when they switch.

  1. VoiceCalls
  2. ChatWhatsApp · web · email
  3. FlowThe logic
  4. KnowledgeGrounded answers
  5. IntegrationsYour systems

One brain

One knowledge base, one set of conversation logic, one memory of the customer.

Whether they called, messaged or emailed, the customer meets the same agent — one that already knows the rest of the conversation.

Flow editor

The ops lead owns the logic. Not the backlog.

Design what the agent asks, when it acts and when it escalates on a visual canvas. Change it at 4pm and it is live at 4:01 — no engineering ticket, no release train.

Renewal dueScheduled
Succeeded
Prepare quotePOST /quotes
Succeeded
Renewals AgentWhatsApp outreach
IFAccepted
See the flow editor

Knowledge

Every answer carries the document it came from

Retrieved from your policy wordings and cited to the clause — not a footnote added afterwards. When the knowledge base cannot support a reply, the agent says so and escalates instead of inventing something plausible.

AI answeredValidated

“Your windscreen excess is SAR 100, and repairs through an approved fitter don’t affect your no-claims discount. A full replacement carries the standard SAR 600 excess.”

Motor Policy Wording v4.2 §7.3POL-1190-A schedule

Grounding is the difference between a confident answer and a correct one.

The citation is produced with the answer and checked against it, so a plausible-sounding wrong answer never reaches the customer.

How grounding works

Under the hood

Understand, retrieve, validate — then act

Calling a model and returning what comes back is a demo. Heard runs every answer through a pipeline, and an answer that fails validation never reaches the customer.

  1. 01

    Understand

    Work out what the customer actually wants, in their dialect, including the half-sentences and the corrections mid-call.

    Intent, entities and sentiment — before a single document is fetched.

  2. 02

    Retrieve

    Pull the specific clause, policy record or order status the answer depends on, from your documents and your live systems.

    Retrieval is scoped to the tenant and the customer. Nothing else is visible.

  3. 03

    Generate

    Compose the reply in the customer’s language and register, with the source it came from attached to it.

    The citation is produced with the answer, not reconstructed afterwards.

  4. 04

    Validate

    Check the answer against what was actually retrieved. If it is not supported, it does not go out — the agent says so and escalates.

    This is the stage most of the category skips. It is the reason the others hallucinate.

Resolved with no human
~66%
Per resolved result · 2 SAR
$0.50
Channels, one agent
5
To a live agent
< 1 day

See Heard on your own calls

A 30-minute call. Bring a recording of a real customer conversation and we will show you what the agent does with it.