A UK voice AI startup is making one of the boldest claims in tech right now: the biggest names in voice AI have been systematically overcharging enterprises. The founder behind it helped scale Vercel into a global giant. Now he says the industry is heading for a major price correction, and he is building the infrastructure to force it.

Voice Labs Have Been Overcharging Enterprises for Years

The voice AI market is exploding. Voice agent usage grew nine times in 2025, and production deployments jumped 340% year-over-year across hundreds of organisations. Gartner predicts conversational AI will cut contact centre labour costs by $80 billion in 2026 alone. But Luke Miller, CEO of UK startup SLNG, says the market has a structural problem built right into it. “The market has been shaped by voice labs whose business model depends on maximising compute at every step of every call,” says Miller. “Every syllable through the most expensive TTS, every pause analysed by a full LLM call, every transcription through the highest-cost engine. We want to reprice the entire voice agent market.” Miller calls this practice “token-maxing”: routing every part of every call through the most expensive frontier models available, regardless of whether the task needs it. Voice labs earn more when compute usage goes up. So the incentive is never to optimise. It is to maximise. **The problem is not just cost. It is outcomes.** Repeatability and reliability matter far more at scale than raw model power on any single call turn. Frontier models called unnecessarily introduce latency, variability, and cost that actively work against enterprise results.

voice AI infrastructure execution layer global sovereign regions

What “The Vercel for Voice Agents” Actually Means

Miller is not building a new AI model. He is not building another voice agent framework. He is building something the industry has been missing entirely: an execution layer. SLNG sits between a team’s existing voice agent orchestrator and the underlying AI models, intelligently coordinating every step of every call in real time. Teams building on frameworks like LiveKit or Pipecat simply plug their existing agent into SLNG. The platform takes over from there. Miller draws a direct parallel to what he witnessed at Vercel, where he was the first seller and built the company’s international business. “Before Vercel, deploying a web app at scale meant stitching together CDNs, build pipelines, edge functions and caching layers yourself,” he says. “Every team reinvented the same infrastructure. Voice agents are at exactly that inflection point.” Here is what SLNG’s execution layer handles across every call:

• Speech-to-text, text-to-speech, and LLM coordination in real time
• Intelligent model routing at every step of the call
• Automatic failover across 11 sovereign regions
• Data residency and compliance enforcement
• Cost optimisation without sacrificing customer outcomes

The platform still calls the best models when a task genuinely demands it. A complex financial advisory question gets treated very differently to a simple appointment confirmation. **The gains come from knowing exactly when a lighter model, or a deterministic response, will deliver a better result than a frontier LLM.** SLNG raised 3.3 million euros in pre-seed funding from Earlybird, StepFunction, and a16z scouts in late 2025 to build this out.

Built Outside Silicon Valley by Design

There is a reason this kind of company could not have been built in Silicon Valley, and Miller is direct about it. In the US, abundant GPU supply encourages teams to throw compute at every problem. In Southeast Asia, the Middle East, Latin America, and India, where much of the world’s enterprise demand actually sits, that option simply does not exist. Co-founder and CPO Ismael Ordaz says that constraint created a different kind of discipline. “When you don’t have abundant GPUs to fall back on, you have to be creative,” says Ordaz. “That discipline now runs through everything we build.” The team built approaches using CPU and memory to handle workloads that competitors route through expensive GPU accelerators. That engineering mindset now powers SLNG’s sovereign infrastructure footprint:

The sectors driving this demand are no surprise. Financial services, banking, insurance, and healthcare are where SLNG’s growth is concentrated, and in these industries, data sovereignty is a legal requirement, not a preference. **”A voice agent handling a mortgage application in Australia cannot have its audio processed in Virginia,”** says Miller. “A patient triage system in Switzerland cannot send recordings to a US-hosted model. These are not edge cases. This is where the enterprise demand is.” Miller also sees Europe’s regulatory environment as a structural advantage. While US voice labs optimise for a single market, SLNG is building distributed infrastructure designed for global compliance from day one.

Real Numbers and a Real Enterprise Customer

The results SLNG is reporting from teams already on the platform are hard to ignore.

• Model costs drop by over 50% across voice agents
• Latency per call turn is cut by more than half
• Target outcomes, including appointment bookings, resolution rates, and conversion, are actually increasing

One early example is Ixigo, India’s second-largest online travel agent. As Ixigo scaled its voice agent operations, the company found itself managing multiple vendor contracts, building in-house voice infrastructure expertise that was not part of its core business, and absorbing the overhead of coordinating it all. Since moving to SLNG, Ixigo accesses all the models it needs through a single platform, has eliminated the vendor management burden, and redirected engineering time toward customer outcomes rather than infrastructure management.

Where Voice Agents Are Heading Next

Miller sees the road ahead clearly, and it goes well beyond cost optimisation. The next phase of voice AI is the agentic economy, where AI agents increasingly build and deploy other agents on demand, and voice becomes the natural human interface for all of it. “We’re positioning SLNG as the default environment for creating voice agents, whether it’s a human building one or an agent spinning one up on the fly,” says Miller. “Voice is how humans interact, and the execution layer needs to be ready for agents to create that interface on demand.” The historical comparison Miller reaches for is deliberate. AWS democratised compute. Stripe democratised payments. Vercel democratised web deployment. **SLNG is betting the execution layer will do the same for voice agents, globally.** For any enterprise already running voice agents at scale, the message from SLNG is straightforward: stop overpaying for compute you do not need, and start optimising for the outcomes you actually want. As voice AI moves from pilot programs to mission-critical infrastructure across industries worldwide, the pricing models and infrastructure decisions made today will define which companies win. That question deserves serious attention right now. What do you think: will voice AI be won on infrastructure, or is the model race still the main event? Drop your thoughts in the comments below.