comutato

// ai sales agent for real estate

$ inbound.$ qualified.> booked()

An AI sales agent for real-estate developers, pre-trained on how properties actually sell in your market. It answers every inbound lead in seconds across WhatsApp, Telegram, Viber and Instagram — qualifying, booking viewings, and logging every conversation to your CRM.

resp.p95
<5s
channels
4
langs
2
handled
>80%
agent.log · tail -f
streamreplaytree
14:02:17[in]whatsapp:+373·6·20** → "hi — is the 2BR on Marina still available for november?"
14:02:17[intent]availability_check · listing=riverfront-1408 · conf=0.97
14:02:18[crm]amocrm.contact.upsert → Ion V. · stage=new_inquiry
14:02:19[out]→ "Yes — unit 1408, €1.2M, river view. I can hold it 24h. Viewing Thu 11:00?"
14:02:34[in]"book me for 11:00"
14:02:35[skill]schedule.viewing(listing=1408, slot=thu-11:00) · conf=0.94
14:02:35[crm]amocrm.stage → viewing_scheduled · value=€1.2M
14:02:36[out]→ "Confirmed. You'll get a reminder. Anything else?"
14:02:40[tick]learning.feedback += 1 · confidence_delta +0.004
14:02:41[wait]listening on 4 channels

/* platform */

Four layers. Shared intelligence.

architecture.txt
channels/
├─ whatsapp · telegram · viber · instagram
│
▼  comutato runtime
│
├─ 04 · learning      feedback loop · per-tenant · zero leakage
├─ 03 · integration   amocrm sync · wazzup24 → kommo
├─ 02 · skills        qualify · schedule · route · hand-off
├─ 01 · language      claude sonnet/haiku · aws bedrock
│
▼
[ inbound agent ]   ·   ru / ro   ·   <5s p95
04
learning layer
Every conversation tunes the system. Per-tenant pattern recognition, zero data leakage.
adaptive · per-tenant
03
integration layer
Two-way AmoCRM sync and omnichannel delivery via wazzup24 → Kommo. Data stays in your systems.
amocrm · wazzup24
02
skills layer
Qualification, scheduling, routing, summaries, hand-off. Typed and adaptable to your playbook.
typed · adaptable
01
language layer
Claude Sonnet & Haiku via AWS Bedrock. Context-aware dialogue in Russian and Romanian.
claude · bedrock

/* workflow */

From contact contract.

$ comutato connect --crm amocrm --channels whatsapp,telegram,viber,instagram
✓ linked 4 channels · pipelines detected
$ comutato train --playbook real-estate --markets cis,romania,dubai
✓ domain playbook ingested · vector.index ready
$ comutato launch --agent inbound --mode supervised
✓ agent listening · responding in <5s p95
$ comutato autonomy --threshold 0.85
✓ >80% of inbound handled without escalation
# every conversation logged with a confidence score
_

/* targets · slo */

text.p95
<5s
response SLA
image.p95
<10s
response SLA
voice.p95
<20s
response SLA
handled
>80%
without escalation

/* faq */

Frequently Asked Questions

Comutato is an AI sales agent for real-estate developers. It answers inbound leads in seconds across WhatsApp, Telegram, Viber and Instagram, qualifies them against your criteria, books viewings, and keeps your CRM clean — pre-trained on how property actually sells in your market.

Four inbound channels — WhatsApp, Telegram, Viber and Instagram — connected through wazzup24 and Kommo. It works in Russian and Romanian today, with more markets and languages as we expand across CIS, Romania and Dubai.

It syncs two-way with AmoCRM — contacts, deal stages, and conversation history stay in your CRM. Comutato operates as a layer on top of your systems, not a replacement.

Every action carries a confidence score. You start fully supervised; as you give feedback, high-confidence actions begin running automatically and anything uncertain is escalated to your team with full context. You stay in control.

Each client runs in an isolated tenant with zero cross-tenant data leakage. Conversations are processed on AWS (Claude via Bedrock) and logged with structured outcomes. GDPR-ready.

// ready()

$ automate(inbound)

Book a demo. See the agent handle your inbound in real time — every lead answered, qualified, and logged with a confidence score.

↩ no credit card · supervised pilot · gdpr-ready