There's a particular kind of inertia that grips enterprise IT departments when someone suggests modernizing the phone system. Not because the technology doesn't exist—cloud contact center platforms have been viable for years—but because the existing Avaya or Cisco UCM deployment, however dated, works. It handles thousands of calls daily. Migrating means vendor negotiations, downtime risk, and months of project management overhead for what amounts to, in many CFOs' eyes, a lateral move.
So most companies don't migrate. They stay put.
Callab AI, a startup that went through Y Combinator's Spring 2026 batch, is betting that this reluctance creates an opening. Instead of asking contact centers to rip out their legacy phone systems and start fresh on a cloud platform, the company built AI voice agents that connect directly to existing on-premises infrastructure using SIP—the protocol that already runs enterprise telephony. The value proposition is disarmingly simple: AI without the migration project.
Whether it's that simple in practice is another question.
Legacy Systems as a Feature, Not a Bug
According to Callab's YC Launch page, posted in late May 2026, the platform integrates natively with legacy PBX systems—Avaya, Cisco UCM, Mitel, FreeSWITCH—and session border controllers. "If it speaks SIP, Callab connects to it," the company says. No middleware, no downtime, no six-month implementation timeline.
The market opportunity, as CEO Haithem Kchaou and CTO Chehir Dhaouadi frame it, hinges on the broader reality that a majority of the global call center industry remains on-premises. The company cites a specific figure—58% of a $400 billion market—though that exact percentage isn't independently verified. Independent research from Metrigy and ContactBabel in 2025–2026 does back up the broader claim that a majority of agent seats worldwide still run on customer-owned infrastructure rather than cloud platforms. The exact percentage is harder to pin down, but the directional point holds. A lot of contact centers haven't moved to the cloud yet. And maybe they never will.
Callab's technical architecture treats that installed base not as a migration challenge but as a distribution channel. Agents register as SIP endpoints on the existing PBX, routing calls without reconfiguring trunk lines or swapping hardware. The company claims it can take an enterprise from contract signature to live AI agents in a week.
One week. That's the pitch, anyway.
What You Get (And What Remains Unproven)
The feature set reads like a checklist of what contact center AI vendors have been racing to deliver over the past 18 months: knowledge base ingestion from PDFs and internal wikis, post-call transcription and sentiment analysis, context-aware transfers to human agents with full conversation history passed along, automatic CRM updates. Callab also emphasizes live supervision—a dashboard where managers can listen to calls in real time, read transcripts as they unfold, and step in if the AI starts to falter.
Under the hood, the company uses a multi-model stack: OpenAI and Groq for language understanding, Deepgram for speech-to-text, ElevenLabs for text-to-speech, according to its sub-processor disclosure page. The architecture supports batch outbound calling at scale—hundreds of concurrent calls, the company says—and integrates with cloud telephony providers like Twilio and Vonage for hybrid deployments where some trunks live in the cloud and others don't.
Callab states it can automate up to 70% of inbound call volume. No independent benchmarks accompany that claim, and the company has yet to publish detailed case studies showing how that plays out in production environments with real customer variability.
The platform also offers a white-label resale model for agencies, BPOs, and systems integrators—a partner motion that lets third parties rebrand Callab's agents and bill usage on their own terms. There's even a "Partner Revenue Calculator" on the company's white-label page, which feels slightly optimistic for a product this early.
Pricing: $500 to Start, With Some Confusion
Callab's pricing, as listed on its site in late May 2026, starts at $500 per month for the Starter tier (5,000 minutes, 20 concurrent calls). Enterprise plans begin at $1,000 monthly with volume-based terms and white-glove onboarding. A one-time setup fee may apply to Starter; Enterprise customers get it waived.
But third-party directory GetApp shows different pricing—Basic at $240/month, Advanced at $600/month—last updated in March 2026. That discrepancy suggests either older tiers that haven't been updated across all channels, or inconsistencies in how distributors list the product. It's the kind of minor confusion that's common in early-stage SaaS but can frustrate buyers trying to budget.
The company lists clients across three continents on its YC profile, including a reference to Dunkin Donuts, though that customer relationship remains unverified in public sources as of late May 2026. This isn't unusual for enterprise software—customers often don't want to be named publicly until they've scaled deployments—but it does leave Callab's track record somewhat opaque.
Crowded Market, Specific Wedge
Callab's launch comes at a moment when contact center AI has become not just viable but expected. Chatbase extended its platform to voice in early May 2026. Regal AI introduced self-improving agents in April. Telnyx demoed real-time voice constraints for AI in February. PolyAI, Replicant, and Parloa have been deploying conversational agents in enterprise contact centers for months, if not longer.
What differentiates Callab, the company argues, is the legacy integration wedge. While many voice AI vendors assume cloud telephony stacks or CPaaS platforms, Callab positions itself on the installed base of on-prem PBX systems that enterprises have been running—and paying maintenance on—for years. Competitors like Cognigy do offer SIP trunking and PSTN connectivity, so Callab isn't alone in this approach. But the messaging centers squarely on zero-migration deployment speed, which is either a genuine differentiator or a marketing angle, depending on how cleanly those one-week timelines hold up in practice.
One thing the company hasn't completed: HIPAA compliance. It's on the roadmap, which makes sense—healthcare and financial services are obvious target verticals for call center AI—but without certification, regulated industries remain off-limits for now. That's a gate, not a roadblock, but it does narrow the early addressable market.
The Team and the Entity
Kchaou spent more than seven years building AI solutions across France, Tunisia, and the UAE before co-founding Callab. Dhaouadi brings a decade of experience in scalable enterprise systems. The legal entity, Clusterlabs Ltd., operates out of Abu Dhabi's financial free zone. YC lists the team size at seven; LinkedIn shows 11–50 employees. That discrepancy likely reflects timing differences, contractors, the parent company structure, or how different platforms count headcount—common enough in distributed startups, but worth noting.
The product is live. The one-week deployment promise is bold, perhaps more so than the founders initially expected. Whether enterprises will trust AI agents to handle 70% of their inbound volume—and whether Callab can deliver on that automation rate consistently—will depend on early production deployments and how quickly the company moves from marketing claims to verifiable case studies.
For now, Callab is making a bet that legacy infrastructure isn't a liability to work around, but an asset to build on. If they're right, the call centers that never migrated to the cloud might leapfrog straight to AI. If they're wrong, well—there are worse problems than having a stable phone system that already works.
