Voice AI revenue just crossed a number that pushes it past “interesting demo” and into “serious enterprise category.” On May 5, ElevenLabs disclosed it hit $500 million in annual recurring revenue. That is up from $350 million at the end of December. Then BlackRock and Nvidia wrote checks. But underneath the headline, three other stories this week point in the same direction. OpenAI opened ChatGPT advertising to any US business. Ashby rewired its applicant tracking system around AI agents. Apple looks set to let you swap Claude or Gemini in for ChatGPT on iPhone. Voice, ads, hiring, OS-level AI. Four surfaces, one pattern. AI providers are turning into utilities.

Voice AI Revenue Hits $500 Million as Audio Becomes a Real B2B Category

ElevenLabs disclosed on May 5 that it crossed $500 million in annual recurring revenue. That is up from $350 million at the end of December 2025. In other words, the company added $150 million in ARR in roughly four months. The Series D round was originally announced at $500 million in February, valuing the company at $11 billion. Since then it has expanded past $550 million. New institutional investors include BlackRock, Wellington, D.E. Shaw, Schroders, Nvidia’s NVentures, and Santander. Sequoia and Andreessen Horowitz remain on the cap table. (Source: TechCrunch, Tech.eu)

Why this voice AI revenue jump matters for HR leaders and founders

First, this is the first AI application category outside the foundation-model labs to scale to nine-figure ARR. Real enterprises are paying real money for real workloads. Nvidia, Salesforce, Santander, KPN and Deutsche Telekom already run voice agents through ElevenLabs. Use cases span customer support, multilingual sales calls, ads, and hiring funnels. For a 50 to 500 person company, the unlock is concrete. Voice screening for high-volume roles, think BDR, support, and hourly retail, is now a buy decision. It used to be a build decision.

Meanwhile, the picks-and-shovels economics matter. AI revenue used to flow upstream to the model labs. However, more of it now flows to vertical specialists. They sit between the model and the application. That is where voice AI revenue compounds. It compounds in the application layer, not the model layer. For founders, the same pattern probably plays out in vision, code, and document understanding over the next 18 months. Specifically, expect AI applicant tracking systems to bundle voice screening as a default feature by Q3.

What to do this week: If you are evaluating voice tooling, run a 30-day pilot against a baseline. The vendor calculus changes when the category leader hits half a billion ARR. Pricing power follows.

OpenAI Opens ChatGPT Ads to Every US Business

On May 5, OpenAI expanded its ChatGPT Ads Manager self-serve beta to all US businesses. It also dropped the $50,000 minimum spend that had gated the original pilot. In addition, OpenAI added cost-per-click bidding alongside its existing CPM model. The recommended starting bid is $3 to $5 per click. A Conversions API and pixel-based measurement tools shipped the same week. (Source: OpenAI, Search Engine Journal)

For HR marketing and talent acquisition teams, ChatGPT is suddenly a paid acquisition channel. Now it sits alongside LinkedIn and Indeed. But the audience there is in a high-intent, problem-solving mode. As a result, a $5 CPC for a senior engineer query behaves very differently from the same CPC on a job board. Your employer brand campaign and your job-ad campaign now compete for the same dollar. Test it small first. Start with a $500 weekly budget on a single role. Then measure application quality before you scale.

Ashby Goes AI-Native at the Data Layer

At its Ashby One conference on May 7, Ashby launched a wave of AI features. These include Custom Agents and the Ashby Assistant chat interface. Slack-embedded agents shipped too. Autonomous Scheduling Agents now handle interview scheduling end to end. Ashby also added Model Context Protocol support. In addition, the company acquired Talent Llama late last year to build an AI Interviewer. Finally, it unveiled People Workflows extending Ashby into onboarding. (Source: PRNewswire, Ashby)

Recruiting tech is now going AI-native at the data layer. It is not just stapling chatbots onto legacy tables. For 50 to 500 person companies on Greenhouse, Lever, or Workday Recruiting, this is a familiar pressure. It is the second time in 18 months that the AI-first challenger has reset expectations. However, an ATS swap is rarely worth the disruption. Instead, push your current vendor on three things this quarter. First, an MCP endpoint for your HRMS. Second, autonomous scheduling. Third, a candidate-debrief agent. If they cannot ship those within two quarters, start a procurement review. The new AI agents for HR wave is where ATS choice becomes a stance on autonomous hiring.

Apple Prepares iOS 27 Extensions, Ending ChatGPT’s Exclusive Slot

Apple is preparing a feature internally called Extensions. It will let users pick Claude, Gemini, Grok, or ChatGPT to handle tasks. The routing covers Siri, Writing Tools, and Image Playground. Reports point to a WWDC announcement on June 8. Shipping is expected in iOS 27, iPadOS 27 and macOS 27 this fall. Gemini stays the default brain via a reported $1 billion-a-year partnership. However, Extensions adds a routing layer on top. (Source: 9to5Mac, MacRumors)

This ends ChatGPT’s exclusive slot inside Apple Intelligence. It also creates fresh procurement questions for IT and security teams. For HR and ops leaders, expect employees to mix AI providers on the same managed device next year. As a result, your BYO-AI policy moves on. It moves from “do we allow ChatGPT on work iPhones” to “which models can touch which data.” Get ahead of it. First, audit which third-party models your company permits for sensitive content this quarter. Then write a one-page policy before the iOS 27 betas ship in July. The top AI tools for HR in your stack should appear on that list. Each one needs explicit data-handling notes.

Quick Hits

• South Korea is refining its AI Basic Act through a 40-plus expert public-private task force. The group is sharpening definitions like “high-impact AI” and compliance pathways. Penalties kick in after the one-year grace period that ends January 2027. (Source: KoreaTechDesk)
• Stanford researchers Echarghaoui, Wu and Fox published BALAR on May 6. It is a Bayesian outer-loop method. The agent picks the next clarifying question by maximising expected mutual information. Benchmarks run on detective cases, thinking puzzles, and clinical diagnosis. (Source: arXiv:2605.05386)
• Delhi-based Kimbal Technologies closed a $22 million Series B led by GEF Capital. The funding scales its Made-in-India smart meters, RF-Mesh modules and AI meter-reading software. Expansion targets include Singapore, Seoul and Australia. (Source: YourStory, Inc42)

If voice AI revenue, agentic ATS launches and OS-level AI routing make you want to rethink your stack, that is the right instinct. Therefore, pick tooling that does not lock you in. Asanify’s HRMS ships with API-first integrations. The AI tools you pick today plug in without a six-month rebuild.

FAQ: Voice AI Revenue, AI Hiring Agents and OS-Level AI

What does the ElevenLabs $500 million ARR milestone tell us about voice AI revenue?

It tells us voice AI is the first AI application category outside the foundation-model labs to reach nine-figure ARR. Real enterprise customers are paying for it. Voice AI revenue is now compounding in the application layer. It sits between the raw model and the end product. For founders, the same playbook will likely repeat across vision, code, and document AI over the next 18 months.

How will OpenAI’s self-serve ChatGPT ads change recruiting?

OpenAI dropped the $50,000 minimum spend. It also added CPC bidding starting at $3 to $5 per click. As a result, any US business can now buy ChatGPT ads. For talent acquisition teams, this is a new high-intent paid surface. It competes with LinkedIn and Indeed budgets. Start with a $500 weekly test on a single role. Then measure application quality and scale.

Should we switch ATS now that Ashby went AI-native with agents?

Probably not yet. ATS migrations are painful. They are rarely worth the disruption mid-cycle. Instead, push your current vendor on three things this quarter. First, an MCP endpoint for your HRMS. Second, autonomous interview scheduling. Third, a candidate-debrief agent. If they cannot ship those within two quarters, start a procurement review.

Not to be considered as tax, legal, financial or HR advice. Regulations change over time so please consult a lawyer, accountant  or Labour Law  expert for specific guidance.