OpenCall.ai and the Quiet Business of Answering the Phone

Call a dermatology office at 5:40 on a Friday and you already know how it goes. It rings. Then it rings some more. Then a voicemail box you will never hear back from. For the practice, that ring is not just an annoyance - it is a patient, an appointment, and a bill that just walked out the door. OpenCall.ai was built for that exact ring.

The San Francisco company, part of Y Combinator's Winter 2024 batch, makes AI voice agents that pick up business phone calls and actually do something with them. Not a phone tree. Not a "press 1 for hours" recording. An agent that holds a conversation, books the appointment, checks the insurance, and updates the record - then hangs up and takes the next call. Its one-line pitch on the YC site is refreshingly flat: "Practice admin run by AI."

That modesty is the interesting part. Plenty of AI startups in 2024 promised to replace whole departments. OpenCall picked the single most repetitive task in a service business and went deep on it.

Section 01The problem hiding in a busy signal

Every service business runs on inbound calls it cannot always answer. Staff are with a patient, on another line, or gone for the day. Industry estimates that OpenCall cites put the leakage at 10 to 30 percent of revenue lost to unanswered or abandoned calls. For a medical practice, that is not a rounding error. It is the difference between a full schedule and a half-empty one.

The old fix was a call center - expensive, hard to staff, and rarely fluent in the specifics of your business. The newer fix was a chatbot, which customers on a phone cannot use. OpenCall's wager is that the phone line itself is the surface that matters, and that an AI good enough to work it is finally here.

There is a reason the phone refuses to die. A patient calling to move an appointment, a customer confirming coverage, an anxious parent asking whether a symptom warrants a visit - these are exactly the interactions people do not want to type into a form. The call is high-intent by definition. Someone dialing your number has already decided they want something. Letting that call ring out is the most expensive way to lose a customer, because you never even learn they tried.

Section 02What the agent actually does

The demo that gets attention is the voice - low latency, natural cadence, no robotic pause before every reply. OpenCall says it runs proprietary models tuned to cut that latency so a caller is not left waiting on dead air. But the founders are clear that the voice is table stakes. The product is the work behind the call.

That last step is the quiet moat. Answering a call is easy. Finishing the multi-step workflow behind it - the insurance verification, the record update, the provider preference, the follow-up text - is where lighter "AI receptionist" tools tend to fall over. OpenCall says its models were trained on thousands of real-world workflows precisely so the awkward middle of a call does not break the agent.

Consider a single appointment change. A human at the front desk pulls up the chart, confirms the patient's identity, checks which provider they usually see, finds an open slot that matches the insurance, moves the booking, and fires off a confirmation text. That is six systems and six decisions inside one two-minute call. An AI that only handles the greeting has automated the easiest ten seconds of that sequence. OpenCall's claim is that it handles the other hundred and ten. Whether it does that reliably at scale is the whole question a buyer is really asking - and the reason the company leans on integrations with tools businesses already run, from CRMs like HubSpot to lightweight databases like Airtable, rather than asking anyone to rip and replace.

"Passionate about music, design, and eliminating tedious work."Arthur Silverstein, co-founder

Section 03Why healthcare first

OpenCall could point its agents at any business with a phone. It chose to lead with medical practices - dermatology, dental, and the multi-location groups and rollups that operate them. Healthcare is a hard beachhead: it demands HIPAA compliance, tolerates almost no error on patient data, and speaks in a dialect of insurance codes and clinical terms. Most voice-AI startups quietly route around it.

OpenCall walked toward it. HIPAA-compliant call handling, insurance jargon parsing, and provider-preference management are core to the pitch, not an enterprise upsell. The logic is familiar to anyone who has watched a wedge strategy work: if you can survive the hardest, most regulated version of the problem, the easier verticals - pest control, retail, e-commerce - are downhill from there.

Section 04The founders and the wager

OpenCall was founded in 2022 by four people: Oliver Silverstein, who serves as CEO; his brother Arthur Silverstein; Nicholas Lee; and Eric Mei. Oliver studied computer science at Northeastern and did stints at Arm and Pluralsight before startups; the team came out of an earlier AI venture, Grey Matter Labs, already circling hard machine-learning problems before narrowing to a single phone line.

Narrowing is the underrated founder skill. It is easy to build a general "AI for business." It is harder, and usually smarter, to give up the other ten good ideas and own one boring task completely. OpenCall's decision to be the thing that answers the phone - and finishes the job - is that kind of focus.

The framing they use matters too. There is a version of AI marketing that promises to replace everyone in the building, and it tends to make the people who actually sign the contract nervous. OpenCall's version is quieter: answer the call the receptionist could not get to, take the after-hours overflow, handle the routine so staff can handle the exceptions. That positioning is less thrilling in a pitch deck. It is also a lot easier for a practice manager to say yes to.

Section 05The market it is racing into

The voice-agent space filled up fast. Retell AI offers no-code agent building. Vocca and CureAgent chase healthcare bookings. Talkie.ai, Emitrr, Gridspace, and a long tail of "AI front desk" tools crowd the same buyer. Against that field, OpenCall's differentiation is not a prettier voice - it is depth of integration and workflow completion, plus the willingness to do the unglamorous compliance work up front.

Answering the call is the demo. Finishing the workflow behind it is the product.

The business model is straightforward B2B SaaS: sell against the cost of human call-center staff, price on the savings, and land-and-expand across a customer's locations. The headline number is the 50 to 80 percent cost cut. The quieter, harder-to-prove number is the revenue that was leaking out through unanswered calls all along - and that one, if OpenCall can show it, is worth far more than the savings.

Section 06Where it stands

OpenCall joined YC in the Winter 2024 batch and has raised early-stage funding, with public sources reporting backing from Y Combinator and CoreNest Capital. The team is still small - around nine people building real-time voice infrastructure that has to work on the first ring, every time. That is a heavy load for a lean team, and it is also the bet: that a focused group can own one task better than a big company owns fifty.

The risks are the obvious ones. Voice AI is a loud, well-funded category, and the gap between a clean demo and a system that never mishears an insurance ID at 6 a.m. is where startups either earn trust or lose it. Regulated healthcare punishes mistakes harder than most markets. And "deploy in under an hour" is a promise that only holds if the integrations behind it stay solid as customers pile on edge cases. None of that is unique to OpenCall - it is the tax on trying to automate something people actually depend on.

The phone call has survived every wave of software that was supposed to kill it. Chat, email, portals, apps - people still call. OpenCall.ai is not trying to end the phone call. It is trying to make sure someone, or something, always answers.