OpenAI on Wednesday introduced GPT-5.2, an update that lets the model plan and execute multi-step tasks across connected applications without step-by-step prompting. The company is positioning the release as its most significant move toward dependable AI agents for the enterprise. In a live demonstration, the model booked travel, reconciled a spreadsheet, and drafted a follow-up brief in a single run, pausing only to confirm irreversible actions. Executives said early access partners reported meaningful reductions in routine operational work. Analysts cautioned that reliability and oversight remain the key questions for agentic systems, but described the launch as a clear signal that 2026 will be defined by AI that acts, not just answers.
CarrierOS