OpenAI pushes specialised agents

OpenAI is turning ChatGPT into a set of job-specific assistants, adding shared workplace agents and a clinician version on April 22. (openai.com) The new workspace agents are available in research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. OpenAI said teams can build shared agents that run in the cloud, connect to approved tools, and keep working across handoffs even when a user is offline. (openai.com) OpenAI’s help documentation says workspace agents can be created, tested, published to a company workspace, used inside Slack, and scheduled to run automatically. At launch, the feature is off by default for ChatGPT Enterprise workspaces, and it is not available for Enterprise customers using enterprise key management, or EKM. (help.openai.com) The company is also widening access to agent mode, the part of ChatGPT that can browse sites, fill forms, edit spreadsheets, and work with connected files and data sources under user supervision. OpenAI’s help center says that mode is built for multi-step online tasks rather than one-off chat answers. (help.openai.com) A separate launch on April 22 pushed the same specialization into healthcare. OpenAI said ChatGPT for Clinicians is now free for verified U.S. physicians, nurse practitioners, physician assistants, and pharmacists, with tools for documentation, medical research, and evidence review. (openai.com) OpenAI said the clinician product includes trusted clinical search, citations, reusable skills, deep research across medical literature, and support for continuing medical education credits on eligible clinical questions. The company’s healthcare help page describes a broader enterprise version, ChatGPT for Healthcare, as a HIPAA-supporting workspace for clinicians, administrators, and researchers. (help.openai.com 1) (help.openai.com 2) The common thread is that OpenAI is packaging AI less as a blank chatbot and more as software for repeatable work. Its Academy materials define workspace agents as systems with a trigger, a process, and approved tools, built to handle recurring tasks that usually require people to re-explain steps and move information between apps. (openai.com) OpenAI is also keeping older products in place while it makes that shift. The company said GPTs will remain available during the rollout of workspace agents, and that it plans to let teams convert GPTs into workspace agents later. (openai.com) In healthcare, OpenAI is pairing the product push with benchmarking. On the same day it launched ChatGPT for Clinicians, the company also introduced HealthBench Professional, which it described as an open benchmark for real clinician chat tasks in care consults, writing and documentation, and medical research. (openai.com) The result is a clearer map of where OpenAI wants ChatGPT to go next: not one assistant for every question, but a growing stack of agents tuned for sales teams, office workflows, and clinical work. The company’s April 22 releases put that strategy into product form. (openai.com 1) (openai.com 2)