Voice AI Industry Expansion

Voice is the next frontier for AI Agents, but most builders struggle to navigate this rapidly evolving ecosystem. After seeing the challenges firsthand, I've created a comprehensive guide to building voice agents in 2024. Three key developments are accelerating this revolution: -> Speech-native models - OpenAI's 60% price cut on their Realtime API last week and Google's Gemini 2.0 Realtime release mark a shift from clunky cascading architectures to fluid, natural interactions -> Reduced complexity - small teams are now building specialized voice agents reaching substantial ARR - from restaurant order-taking to sales qualification -> Mature infrastructure - new developer platforms handle the hard parts (latency, error handling, conversation management), letting builders focus on unique experiences For the first time, we have god-like AI systems that truly converse like humans. For builders, this moment is huge. Unlike web or mobile development, voice AI is still being defined—offering fertile ground for those who understand both the technical stack and real-world use cases. With voice agents that can be interrupted and can handle emotional context, we’re leaving behind the era of rule-based, rigid experiences and ushering in a future where AI feels truly conversational. This toolkit breaks down: -> Foundation layers (speech-to-text, text-to-speech) -> Voice AI middleware (speech-to-speech models, agent frameworks) -> End-to-end platforms -> Evaluation tools and best practices Plus, a detailed framework for choosing between full-stack platforms vs. custom builds based on your latency, cost, and control requirements. Post with the full list of packages and tools as well as my framework for choosing your voice agent architecture https://lnkd.in/g9ebbfX3 Also available as a NotebookLM-powered podcast episode. Go build. P.S. I plan to publish concrete guides so follow here and subscribe to my newsletter.

IBM just picked its first-ever voice AI partner, and it's not who you'd expect. February wrapped with enterprise voice AI moving from experiments to infrastructure. The deals this month aren't proofs of concept. They're production integrations with real revenue attached. Deepgram lands inside IBM watsonx. IBM selected Deepgram as its first voice partner, embedding STT and TTS directly into watsonx Orchestrate. Enterprise-grade transcription, real-time captioning, and multilingual support are now native to IBM's AI platform. This is voice AI graduating from a standalone tool to an embedded enterprise layer. Anthropic goes all-in on enterprise agents. Claude Cowork shipped private plugin marketplaces, 10 department-specific plugins (finance, legal, HR), 12 new MCP connectors (Gmail, DocuSign, Clay), and cross-app workflows for Excel and PowerPoint. Enterprise admins can now curate which AI capabilities their teams access. OpenAI drops GPT-Realtime-1.5. The updated real-time voice model brings +10% transcription accuracy, +7% instruction compliance, and +5% audio reasoning at the same price. Connection success rates climbing to ~66% and error rates cut in half suggest this model is getting production-ready. SoundHound AI AI takes voice agents to the retail floor. Sales Assist, unveiled at MWC 2026, listens to in-store customer conversations and pushes real-time deal recommendations to staff devices. Built on SoundHound's Polaris ASR, purpose-built for noisy environments. ElevenLabs closes $500M at $11B. Sequoia led, a16z quadrupled down. With $200M+ ARR and expansion into multimodal agents and 14 global offices, ElevenLabs is building toward an IPO. Newo.ai raises $25M for AI receptionists. Their zero-hallucination architecture runs parallel AI agents to verify responses before they reach callers. 15,000+ agents deployed, mostly in dental, restaurants, and home services. Speechmatics signs two enterprise deals in one week. Partnered with VCONIC for healthcare/financial compliance and Boost AI for regulated European industries. Their medical model hits 93% real-world accuracy with 50% fewer terminology errors. Voice AI isn't waiting for permission anymore. Which of these moves matters most to your stack?

🧵 This week in conversational AI: This week reinforced a clear theme: Voice AI is entering its scale phase, where reliability, latency, and control really matter. Here’s the recap 👇 Deepgram sees its latest funding highlighted by The Wall Street Journal, valuing the company at $1.3B. Real-time voice APIs are officially core infrastructure. ElevenLabs drops 𝗦𝗰𝗿𝗶𝗯𝗲 𝘃𝟮 + 𝗦𝗰𝗿𝗶𝗯𝗲 𝘃𝟮 𝗥𝗲𝗮𝗹𝘁𝗶𝗺𝗲, delivering sub-150ms transcription across 90 languages with ~93%+ accuracy. This is the latency threshold where voice stops feeling like software and starts feeling human. VoiceRun raises a $5.5M seed and launches a full-stack, code-first Voice AI platform for enterprises. Control, observability, and reliability are becoming non-negotiable as voice agents graduate to production. OpenAI releases “𝘈𝘐 𝘢𝘴 𝘢 𝘏𝘦𝘢𝘭𝘵𝘩𝘤𝘢𝘳𝘦 𝘈𝘭𝘭𝘺,” showing how millions of Americans are already using ChatGPT to navigate a broken healthcare system. Conversational AI is emerging as a critical layer for access, clarity, and patient empowerment. Parloa announces a $350M Series D at a $3B valuation, just seven months after its Series C, led by General Catalyst. The company is accelerating global growth, expanding its AI Agent Management Platform, and launching the Parloa Promise, a strong signal that enterprise-grade, responsible AI is scaling fast. Krisp launches webhooks for its AI Meeting Assistant, letting transcripts, notes, and action items flow directly into internal tools. Voice → structured data → action, without friction. NVIDIA releases Nemotron Speech ASR, an open-source model hitting ~24ms median transcription time with massive concurrency on H100s. Real-time voice at scale just became far more accessible. SoundHound AI x Richtech Robotics partner to bring conversational voice AI into robotic food service. Voice continues to emerge as the interface between humans, machines, and real-world transactions. 🚀 Big week for conversational AI. What did we miss?

AI’s full takeover of the Call Center is now more of a rollout and change management problem than anything else. Speech quality, agentic orchestration, and compliance are continuously improving and are on track for the takeover scenario. If procurement cycles and change-management hurdles don’t slow things down, my guess is that Tier-1 voice support should be 90% automated by 2030, with humans fully shifting to AI training, coaching and exception handling roles by 2032 (also AI assisted). For highly regulated industries like banks, insurers, and telecoms, they’ll likely choose a two-layer strategy: hyperscaler CCaaS for the spine, plus a specialized voice-bot vendor for high-stakes domains (fraud, collections) until confidence, cost, and regulation catch up. Here’s what needs to be true for this all to happen: First, you need an ultra-reliable voice stack that’s <300 ms bidirectional latency so conversations feel human. WER <3% across dialects and background noise (the latest speech models set the bar here). Agentic orchestration, not just intent detection, has to be available as models must engage backend systems (think CRM, payments, logistics) safely and independently. Also, multi-agent planners (like those announced in Microsoft Copilot Studio in 2025) can deliver on the architectural path. Third, the right guardrails to deliver retrieval-augmented generation tied to authoritative knowledge bases; proofs of source logged for compliance. Real-time redaction and PII masking baked into the pipeline to satisfy HIPAA, PCI-DSS, and any emerging AI policy requirements. Cost parity will be another key ingredient. Inference plus carrier fees needs to stay below what the fully-loaded labor cost of an offshore agent is in 2025. Onshore comes later. Finally, you need supervisor copilots, quality-assurance bots like Genesys’ “AI for Supervisors” or NICE’s Enlighten Copilot, and continuous training loops to replace the traditional floor manager model. This is already happening, so it’s more a matter of time to see the capabilities improve. Gartner still expects only ~10% of all agent interactions to be automated by 2026, up from 1.8% in 2022. So the gap between achievable tech and enterprise rollout is the real drag on the timeline. The tech is largely “here.” The reason you aren’t seeing +80% voice automation in the average contact centre yet is mostly enterprise readiness. Think legacy systems, undocumented know-how, risk governance, and slow org redesign. Yes, the tech still needs polish in multi-step reasoning and compliance, but the heavier lift right now is inside the enterprise walls, not the model weights. If you’re in a call center role today, how are you approaching this? #ccaas #contactcenter #ai

Voice AI is a labor category. It competes with payroll, not software features. The commercial maturity data makes that clear. The category has moved past “cool demo,” with 79% of private Voice AI companies now deploying or scaling. Not because they’re getting better at talking, but because they’re getting better at doing work. Business relationship data highlights three clusters where Voice AI is seeing the most traction: 1) The new voice of media and entertainment Voice AI is being used for dubbing, localization, narration, accessibility, and content production. Work that used to require large teams and long turnaround times. 2) High-volume service industries are where ROI shows up fastest Healthcare, financial services, contact centers, hospitality. Environments where calls are repetitive, demand is constant, performance is measured in minutes per interaction, and cost is measured in labor hours and cost-per-resolution. 3) Real-time voice raises the engineering bar and drives infrastructure partnerships These systems operate in live customer interactions. Latency, uptime, error rates, monitoring, and integrations matter more than “how smart the model sounds,” because failures happen directly in front of customers. The most successful deployments are already replacing or reducing paid conversational work: agents, schedulers, operators, coordinators. They connect directly to the systems those workers use. Adoption is being led by industries where voice equals cost, capacity, and time-to-resolution. Where do you think Voice AI replaces the most labor over the next 3 to 5 years? Where will we still talk to humans? P.S. More on this in CB Insights 2026 Tech Trends | cc Isabelle Lowe

#VoiceAI just crossed a line most of us didn’t see coming. Alibaba’s #Qwen3-TTS-1.7B isn’t another “better robot voice.” It sounds… human. Uncomfortably so. Natural tone. Emotional range. Accent control. And it runs in real time on everyday hardware. This isn’t a lab demo locked behind enterprise pricing. It’s fully open-source. Real-time. Usable. What stands out isn’t just the feature list, but what it signals. With a few seconds of reference audio, a voice can be recreated. Emotion is no longer implied; it’s instructed. Latency is low enough for live conversations. Languages are handled with consistency, not patchwork fixes. And the license removes the meter that used to tick with every word spoken. The quiet shock is this: Benchmarks show speaker similarity that rivals, and in some cases exceeds, well-known proprietary voice platforms—on a single GPU. That changes the economics overnight. Voice once meant studios, contracts, and per-minute costs. Now it means open models, local deployment, and fully owned voice systems. For builders, this opens doors that were previously bolted shut: Real-time agents that don’t sound synthetic. Accessibility tools that feel respectful, not mechanical. Learning, gaming, storytelling, and support systems where voice is no longer the bottleneck. The interface just became more human. And that’s exactly where the unease begins. When voices can be copied this easily, sound loses its authority. Audio can no longer stand alone as proof. Impersonation, fraud, and social engineering don’t need better scripts anymore. They just need a familiar voice. This is why risk, verification, and trust systems can no longer be optional layers. They are fast becoming core infrastructure. We are stepping into a phase where: Seeing was already questionable. Now hearing is too. Technology taught machines how to speak with us. The harder task ahead is teaching ourselves how to listen—carefully, critically, and with context. Progress didn’t slow down. It just got a voice.

This week on Riding Unicorns, we’re joined by Nikola Mrkšić, Co-Founder & CEO of PolyAI, one of the world’s leading voice AI businesses. Most people still think voice AI is about reducing call centre costs. That misses the point. PolyAI is not just automating calls. It is turning the contact centre into an intelligence layer for the entire enterprise. PolyAI raised $86 million in Series D funding in December 2025, led by Georgian, Hedosophia and Khosla Ventures, bringing total funding to more than $200 million. Also a shout out to Passion Capital and Amadeus Capital Partners, who backed the business at Seed. Nikola was part of the early team behind Apple’s Siri. The vision was right. The timing was not. Now, with better models, infrastructure and enterprise demand, voice AI is finally having its moment. We discuss: • Why PolyAI focused on the gap between traditional Interactive Voice Response (IVR) and truly conversational AI • How restaurants using PolyAI never miss a booking call, with some seeing revenue increase by 5 to 10% • How one customer used PolyAI to handle nearly 1 million calls during a major disruption, making sure customers got immediate information when it mattered most • Why retailers rely on voice AI during peak periods like the holiday season, when call volumes can jump 3 to 5x and human teams cannot scale fast enough • How contact centres can become an intelligence layer for the wider business, surfacing issues like billing errors or depot problems before they turn into larger operational failures • Why enterprise deployments become hard to rip out once the product is deeply integrated across fragmented internal systems • Why Nikola believes many AI businesses are effectively reselling someone else’s models, and why owning both the models and the application layer gives PolyAI greater autonomy and healthier margins • Why outcome-based pricing can look attractive at first, but often breaks down over time, while transparent pricing and real product value create more durable businesses • Why Gordon Ramsay is both a customer and an unusually strong brand fit for PolyAI If you want to understand where conversational AI is actually going, beyond the hype, this is a great episode. 🎧 Listen here: Spotify: https://lnkd.in/e-ZfCkpk Apple: https://lnkd.in/ef_AirAN

OpenAI just reorganized their entire audio team. Not a small tweak. A complete merger of engineering, product, and research divisions over the past two months. Their goal: launch an audio-first device within a year. Here's why this matters for every AI leader right now. Voice assistant users in the US will hit 157.1 million by 2026. But here's the disconnect: less than 20% of users say voice is the easiest way to interact with AI tools. Screens still win. 28-35% prefer touch, 18-35% prefer keyboard and mouse. So why is OpenAI betting everything on audio? Because they're not building for today's preferences. They're building for the moment when voice response times drop below 300ms, the human neurological threshold for natural conversation. Their new model, launching Q1 2026, will handle interruptions and speak while you're talking. Not taking turns. Actual conversation. For enterprise leaders, this creates a decision point. Voice isn't replacing screens. But companies using agentic voice AI are seeing 60% automation of repetitive workflows and faster onboarding. The question isn't whether to adopt voice interfaces. It's whether you're designing for multimodal interaction now, before your competitors force your hand. What's your take? Are you building for voice-first, screen-first, or hybrid?

Every now and then people keep asking me why does India have so many voice AI companies but not as much in other categories? The answer is actually pretty simple. Voice AI is booming in India because domestic demand is booming. Indian companies across every stage are buying voice AI, deploying it, iterating on it, so founders flock there. Rational signal-following. That’s how healthy startup ecosystems form. If we zoom out we’ll also see that Indian founders are also quietly building world-class AI infrastructure for US enterprise - tooling, orchestration, middleware. They’re not absent from B2B AI. They just followed the market that was pulling hardest. And they’ve been winning there. The real question is: what happens when Indian enterprise starts pulling too? We’re starting to see early signs - pockets of genuine willingness to pay in fintech, healthcare, logistics. Not uniform, but directionally real. I don’t think the constraint was founder quality or technical depth. It has been buyer readiness. That’s how markets work. The categories where we haven’t seen as many Indian AI startups? Look closer and you’ll find the same pattern - enterprise B2B buyers in India haven’t adopted those tools yet, so the addressable market looks thin, so founders don’t bite. They’re rational actors. The next wave of Indian AI companies won’t just be voice. It will be whatever category Indian enterprises decide to buy first and back with real budgets. I think agentic automation is the next big one. My bet is we are 6-12 months away from it before this conversation gets really interesting.

I've been experimenting in the Voice AI space for over a year now. Shipped a couple of tools. Had real customers. Learned a ton. For the longest time, all the noise, hype, and cutting-edge infra came from Silicon Valley startups. But something shifted this week. Karan Goel from Cartesia was in India. Multiple Voice AI builder meetups happened in Bangalore. The energy was palpable. Blue Machines AI got their agent to talk to Arnab Goswami for over an hour in the mainstream media. I've watched two startups grow from scrappy beta products to raising millions - Bolna (YC F25) and Ringg AI. What did they get right early on? → Exceptional developer tooling → Well-documented APIs → Seamless integrations with Indian phone numbers (something even well-funded global platforms struggled with) Fun fact: I got my first Voice AI prompts from Ringg. Their prompting for Hindi agents was that good even in Jan 2025. A space that was invisible to Indian VCs just 18 months ago is now a hot cake. The Indian Voice AI ecosystem is thriving: • Ringg AI, Bolna (YC F25), Nurix, Blue Machines AI - Horizontal platforms • Pype AI, Confido Health - Healthcare • Cekura - Evals • Hunar.AI - Frontline Recruitment • Userology - Market Research • smallest.ai - TTS & Infra What excites me most? These aren't copy-paste solutions. Some of them are solving for India-first problems - multilingual support, local telecom infra, cost structures that make sense for emerging markets, while others are building for the world from India. Such exciting times for voice-first platforms.