OpenAI has hired Peter Steinberger to lead its personal agents initiative, a significant talent grab that signals the company's intent to move beyond chatbots toward AI systems that actually do things on behalf of users.
The hiring, confirmed Monday, puts Steinberger in charge of what OpenAI internally calls agents that live in your devices. Steinberger was former technical lead at Apple's Siri team and most recently a director at Google working on Assistant infrastructure.
Personal agents represent the next frontier for consumer AI. While GPT-5 can answer questions and write code, personal agents are designed to take action: booking flights, managing calendars, filling out forms, and handling the administrative overhead of daily life.
Steinberger's background makes him uniquely suited for this. At Apple, he worked on Siri's intent recognition systems—the part that figures out what you actually want when you say remind me to call Mom. At Google, he led infrastructure teams that processed billions of Assistant requests daily.
OpenAI isn't alone in this race. Anthropic has been quietly building agent capabilities into Claude Opus 4.6. Google is expected to integrate agent features into Gemini later this year. And a crop of startups are chasing the same vision.
What makes Steinberger's hire notable is timing. OpenAI has the models. It has the distribution. What it hasn't had—until now—is deep expertise in the messy, integration-heavy work of connecting AI to the systems people actually use.
Sources familiar with the hiring say Steinberger's mandate is to build agents that work across devices and platforms. The goal isn't a better chatbot. It's an AI that can operate your phone, your laptop, and your web services with the same fluency a human assistant would bring.
This requires solving hard problems. Authentication across services. Permission management. Error handling when things go wrong. And perhaps most difficult: building user trust that the agent won't make expensive mistakes.
The hire comes as OpenAI faces pressure to demonstrate practical applications of its technology. GPT-5 is impressive. But investors, enterprise customers, and regulators are all asking the same question: what can it actually do for people?
Personal agents are a plausible answer. Unlike coding assistants or writing tools, agents address a universal pain point: administrative work. Everyone has forms to fill, appointments to schedule, and emails to manage. An AI that handles this reliably would find immediate product-market fit.
The question is whether Steinberger can deliver. Personal agents have been a perennially near-term technology since the 1990s. Siri, Alexa, and Google Assistant all promised variations on this vision. None fully delivered.
OpenAI is betting that foundation models change the equation. Where previous generations of assistants relied on hand-crafted rules and limited domains, GPT-class models can handle open-ended tasks with genuine comprehension.
Steinberger starts immediately. His first priority is building the infrastructure layer—the authentication, permission, and error-handling systems that any viable agent requires.
Product integration will follow. OpenAI hasn't announced a timeline, but the hiring suggests significant agent capabilities could appear in ChatGPT or standalone products within the year.
