OpenAI is building AI agents for everything. Will everyone use them? | TechCrunch

OpenAI is attempting to expand the reach of its artificial intelligence agents beyond software developers to the general white-collar workforce through its new ChatGPT Work service.

This product, offered at a monthly subscription rate, is designed to automate complex, multi-step digital tasks for professionals like accountants and marketers by granting the underlying language models access to everyday work tools such as email, calendar apps, and communication channels. While software engineers have readily embraced agentic tools for coding, OpenAI faces the challenge of convincing mainstream office workers to trust these autonomous systems with their sensitive digital workflows and personal accounts.

The transition from developer-centric tools to general-purpose business assistants introduces several product and technical hurdles, including complicated permission setups, high computational token costs, and the difficulty of evaluating non-code outcomes. To bridge this gap, OpenAI's design team is incorporating temporary user-interface features like buttons to guide non-technical users, even as internal purists argue that advanced models will eventually handle everything through simple text prompts. As OpenAI competes with rivals like Anthropic to establish its platform as the primary digital workspace assistant, the company relies heavily on the hope that rapid improvements in model efficiency and capability will solve current usability limitations before users abandon the technology.