How Tunisian-born Clusterlab is Making Voice AI Smarter for the Region

Clusterlab, a Tunisian-born startup headquartered in Dubai, builds Callab.ai — voice AI agents tailored to Arabic dialects and legacy telephony — and was admitted to Y Combinator's P26 seed batch. The company targets contact-centre automation across industries in the Middle East.

Tunisian-born, UAE-headquartered startup Clusterlab is aiming to close a major gap in regional voice artificial intelligence with its product Callab.ai. Founded in 2020 by CEO Haithem Kchaou and CTO Chehir Dhaouadi, the Dubai-based company builds voice AI agents tailored to Arabic and its regional dialects while integrating with legacy telephony systems. Clusterlab’s acceptance into Y Combinator’s P26 batch — the seed-stage accelerator known for alumni such as Airbnb and Stripe — marks a significant credibility boost as the firm scales its contact-centre automation across industries. Industry research underlines the opportunity: a 2024 Gartner report projects that at least 70% of customer service journeys will begin and be resolved through conversational GenAI or third-party assistants by 2028.

“The global AI industry was building in English, for Western users, on Western data, and the Middle East was expected to adapt,” Kchaou explains. “We wanted to build the opposite, AI designed from day one for the languages, dialects, and infrastructure of our region.”

How Callab.ai works

Callab.ai is presented as a voice AI agent that can answer calls, understand callers in their own language and dialect, and either resolve requests end-to-end or transfer them to human agents with full context preserved. Kchaou cites an early project, Elm — an AI-powered library developed for Tunisia’s Ministry of Higher Education — as a test case that exposed how global AI tools often fall short outside Western markets. That experience informed Clusterlab’s strategy to prioritise three core gaps: language, data and infrastructure.

“Callab.ai was built from the integration layer upward. We started with the hardest part, native compatibility with the telephony stacks enterprises already operate, and then built the AI capabilities on top of that foundation,” Kchaou explains, noting that typical contact centres in the region still run on platforms such as Cisco Unified Communications Manager (CUCM) or legacy Private Branch Exchange (PBX) systems.

  • Founders: Haithem Kchaou (CEO) and Chehir Dhaouadi (CTO)
  • Founded: 2020; HQ: Dubai, UAE
  • Product: Callab.ai — multilingual voice AI with emphasis on Arabic dialects
  • Notable project: Elm for Tunisia’s Ministry of Higher Education
  • Y Combinator: Admitted to P26 batch; described as the first company with Tunisian roots in YC’s history by Kchaou

Clusterlab positions Callab.ai as a 24/7 voice agent across sectors including real estate, healthcare, hospitality, retail and debt collection. The system handles routine tasks such as scheduling, reminders and basic inquiries, with CTO Chehir Dhaouadi emphasising user experience: “No hold music, no IVR menu maze, and a conversation that feels natural rather than scripted.”

Kchaou says the YC programme — a three-month accelerator — is expected to help the company sharpen product-market fit and access longer-horizon capital needed for deep-tech infrastructure. “Being accepted into YC was, first and foremost, an additional confirmation that our region produces world-class solutions and world-class talent. We are proud to be among the small group of UAE-headquartered startups to join YC, and also the first company with Tunisian roots in YC’s history,” he shares.

Looking ahead, Kchaou predicts rapid adoption of voice AI across the Middle East but stresses the region cannot simply import Western models. “What makes our region distinctive is that this transition cannot simply be a copy-paste of what is happening in the United States,” he clarifies, pointing to local data residency requirements, the diversity of Arabic dialects, and entrenched telephony infrastructure as constraints Callab.ai is built to address.