There are 3 million call center agents in the United States alone. Most of their calls follow a script — same opening, same objection handling, same CRM entry at the end. The conversation varies slightly. The structure never does.
That is not a human problem. That is an infrastructure problem. And infrastructure problems have infrastructure solutions.
The market is enormous and largely untouched by real automation
The global contact center market was valued at $496 billion in 2023 and is projected to reach $741 billion by 2030, growing at 6% CAGR. Grand View Research puts this as one of the largest operational expenditure categories for mid-to-large enterprises worldwide.
In the US specifically, outbound calling is the backbone of three industries we are targeting first:
- Debt collections — a $17.8 billion industry (IBISWorld, 2024) employing ~120,000 agents whose primary job is to make outbound calls, negotiate balances, and log outcomes.
- Lead qualification — SDR salaries average $55,000–$75,000/year in the US. Most of an SDR's time is spent on calls that go nowhere. Salesforce's State of Sales report found that sales reps spend only 28% of their week actually selling.
- Appointment booking — healthcare, home services, and financial services collectively lose billions annually to no-shows and manual scheduling overhead. The global appointment scheduling software market alone is projected at $633 million by 2025.
The real cost of a human on the phone
A fully loaded call center agent in the US costs between $25 and $65 per hour when you factor in salary, benefits, training, supervision, and attrition. Average annual agent turnover in call centers runs at 30–45% — meaning companies are constantly hiring and retraining.
The average agent handles 50–80 outbound calls per day. Of those, a significant portion reach voicemail, wrong numbers, or are declined immediately. The productive call rate — calls that result in a meaningful outcome — is often below 20%.
Gartner projected that conversational AI in contact centers would reduce agent labor costs by $80 billion by 2026. That report was published in 2022. We are in 2026. Most of that reduction has not happened yet because the tools that existed were infrastructure — APIs, not products.
What exists today is not enough
The current landscape splits into two camps. On one side: legacy dialers and CRM call tools (Five9, RingCentral, Genesys) that automate the mechanical parts of calling but still require humans to actually talk. On the other: AI voice infrastructure companies like Vapi, Bland AI, and Retell that give developers the primitives to build AI calling — but require significant engineering to turn into a working product.
Neither camp has built what businesses actually need: a system you can point at a use case, configure once, and deploy. Something that calls, converses, handles objections, logs outcomes, and syncs to your CRM — without an engineering team maintaining it.
That is what Voxa is.
Why voice specifically
Text-based automation — emails, chatbots, SMS — has been commoditized. Businesses have learned to ignore it, and recipients have learned to filter it. Voice is different. A phone call still commands attention in a way that a message does not. The conversion rates on a real conversation, even an AI one, are materially higher.
More importantly: the workflows that live in outbound calling are among the last major enterprise workflows that have not been automated at the conversation layer. Email sequences exist. Ad targeting exists. Outbound voice — at scale, with full conversational ability — is still largely a human job in 2026.
That gap is where Voxa sits.
What we are building toward
Voxa starts with outbound calls across collections, lead qualification, and appointment booking. But the underlying infrastructure — a voice agent that can hold a real conversation, handle objections, and take action in a CRM — applies to any workflow that currently requires a human on the phone.
Inbound support. Insurance claims intake. Patient follow-ups. Survey collection. Every one of these is a version of the same problem: a structured conversation that needs to happen at scale, reliably, without a human in the loop.
We are building the first artifact in this space. Not a tool. Not an API. A system that you plug in and walk away from.