Diwan is the Arabic-first autonomous customer-lifecycle OS — a speech-native multi-agent workforce that acquires, onboards, supports, retains, and grows your customers. We don't digitize the call center. We own the relationship.
Everyone is optimizing the cheapest minute of the customer relationship — the support call. We built the engine that owns the other 99%.
The contact center is the last place value is created, and the first to commoditize. Arabic ASR accuracy — today's battleground — is a depreciating asset as sovereign foundation models mature.
So we don't compete on the model. We rent the best Arabic models and build the moat above them: outcome-labeled dialect data, a memory that follows the customer across their whole life with a brand, and in-Kingdom residency a foreign platform can't replicate for years.
Most platforms answer questions. Diwan runs the entire arc — and remembers every conversation, sentiment, and promise as the customer moves through it.
Outbound qualification, lead reactivation, and voice-note replies where MENA commerce actually happens — WhatsApp and Instagram DMs.
Proactive, guided activation that drives customers to first value — instead of leaving it to a help article they'll never read.
Resolution, not deflection. The agent takes real actions in your CRM, payments, and logistics — bounded by your policies.
Detects churn risk from sentiment and behavior, then autonomously wins customers back — the first stage we ship.
Upsell and cross-sell at the right emotional moment in a live conversation — a moment a cost-center tool never sees.
Manages the WhatsApp and Telegram communities that are the social fabric of MENA brands — at scale, in dialect.
We own the two layers that get harder to copy every conversation — the orchestration intelligence and the outcome-labeled data flywheel — and rent the one that gets cheaper every quarter.
Not the old ASR→LLM→TTS pipeline that throws away dialect and emotion. Diwan operates on a joint acoustic-semantic representation, so it hears a frustrated Khaleeji caller differently from an impatient Cairene founder — with sub-second turn-taking and barge-in.
Easy, high-volume turns go to a small in-region Arabic model — cheap, fast, private. Only the hard reasoning turns hit a frontier model. Most turns cost cents. That's the unit-economics engine behind outcome-based pricing.
A persistent, per-customer memory: every prior conversation, sentiment trajectory, purchase, promise made, dialect and tone preference. It's what makes the agent feel like it knows you — and the asset no foundation model has.
Runs in-region on sovereign cloud, with data residency, audit trails, and breach alerting built in — not bolted on. A structural advantage foreign agentic platforms cannot match quickly.
Arabic is morphologically rich and diglossic — MSA plus a dozen dialects. Outcome-labeled dialectal conversation data at scale is something neither global players (no dialect data) nor sovereign model labs (no outcome data) possess. Every conversation widens the gap.
In-Kingdom processing, PDPL & SDAIA-native architecture. A multi-year structural advantage over US/EU entrants — the kind of moat that's boring to build and very hard to cross.
Once Diwan holds a brand's entire customer-relationship memory and is wired into its action systems, ripping it out means amnesia. The Salesforce lesson, applied to voice.
A self-serve layer seeds data and word-of-mouth; outcome wins compound into reference power. Being the Arabic category-definer is a brand moat — winnable exactly once.
Saudi PDPL is enforced by SDAIA with localization requirements and fines up to SAR 5M. The UAE PDPL's executive regulations are now live. We treat residency as a product, not a checkbox.
We're onboarding a small cohort of KSA & UAE design partners in e-commerce, fintech, and q-commerce. If churn, activation, or retention is a number you own — let's talk.
Design-partner program · Arabic-first · in-region by default.