OpenAI launches Workspace Agents: capable of continuous cloud execution with memory, ushering in the era of enterprise digital colleagues

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯1319 words

OpenAI announced the replacement of enterprise Custom GPTs with Workspace Agents, leveraging the Codex code execution foundation to enable AI agents to autonomously complete tasks across over 90 enterprise applications, including Slack, Google Drive, and Salesforce. (Context: OpenAI is rumored to be eyeing an IPO as early as the end of this year with a valuation hitting $1 trillion; ChatGPT must secure more enterprise clients.) (Background: OpenAI is rumored to launch an AI-powered Office suite, setting the stage for a direct confrontation with Microsoft and Google.) OpenAI officially launched "Codex as the true underlying foundation" last night (22nd). Workspace Agents are positioned as the successor to Custom GPTs and are available for ChatGPT Business, Enterprise, Edu, and Teachers plans. OpenAI also announced that it will officially deprecate organizational Custom GPTs at an unspecified future date. The core of this release is the execution foundation behind Workspace Agents: Codex. Codex is OpenAI's cloud-based code execution environment, an infrastructure capable of running code, reading and writing files, calling APIs, and maintaining memory. Six days ago, the Codex desktop version added over 90 plugins, covering tools like Atlassian Rovo, CircleCI, GitLab, and Microsoft Suite, alongside capabilities such as background computer usage, persistent memory, and scheduled wake-ups. Workspace Agents directly inherit this underlying layer. When an agent is tasked with generating a metrics report every Friday, it effectively initiates a session in the Codex cloud environment, calls connected tools, fetches and organizes data, generates charts, writes summaries, and finally saves the results for use the following week. Once deployed to a Slack channel, the same agent acts as a Codex routine that continuously monitors mentions and reports progress within the thread. Not just conversation, but continuous operation The fundamental limitation of legacy AI assistants was that "progress stopped as soon as the user left." Workspace Agents change this: agents run continuously in the cloud, supporting long-term workflows spanning days or even weeks, and can be configured with scheduled triggers. OpenAI's release materials showcased several types of typical agents: - Spark: Responsible for lead qualification and follow-up. - Tally: Periodically generates metrics reports. - Scout: Aggregates product feedback in the #user-insights channel and routes it to relevant teams. - Another accounting agent: Handles parts of the month-end closing process, including journal entries, balance sheet reconciliation, and variance analysis, while retaining working papers for human review. Ankur Bhatt, AI Engineering Lead at Rippling, described a more compelling use case: "It researches customer accounts, summarizes Gong call recordings, and posts transaction summaries directly to the sales team's Slack channel. What used to take sales reps 5 to 6 hours of work per week is now automatically completed in the background for every transaction." Memory is also key. Agents can retain memory across multiple executions, continuously improving through conversational feedback, and gradually reflecting a team's actual working methods, standards, and conventions. This is something static instruction-based Custom GPTs could not achieve. The company stated that Workspace Agents will be free for all eligible plans until May 6; thereafter, it will transition to a credit-based billing model.

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching6 items

2026-04-22

OpenAI Launches Workspace Agents Feature in ChatGPT

2026-04-23

Naval, author of The Almanack of Naval Ravikant, launches AI interval fund USVC, allowing participation in early-stage investments in xAI and OpenAI with as little as $500

Similarity 120%關鍵字 openai同分類 zh

2026-04-23

OpenAI CEO Sam Altman slams Anthropic: Fear-mongering Claude Mythos is just to monopolize AI

Similarity 120%關鍵字 openai同分類 zh

2026-04-23

OpenAI launches ChatGPT for Clinicians, assisting human physicians with 99.6% accuracy (free access in the US)