Can an AI agent complete a full day's office task unattended and report back accurately?
YesOpenAI says its deployed research agent completes well-defined assignments that would take a skilled person several days and is now used extensively across its research organization.
From the Editors
Some tools reported to do this
Grok Bot ↗OpenAI automated research intern ↗Cursor Projects ↗Salesforce Agentforce ↗Claude Code Projects ↗Sourcegraph Agentic Batch Changes ↗ChatGPT Work ↗ChatGPT Scheduled Tasks ↗ChatGPT agent ↗
Related Solution · Sponsored
Reach readers asking what AI can do for them in your category.Advertise here
Some sources
The Threshold Report: OpenAI agents complete multi-day research tasks · thethreshold.report
Research acceleration view inside openai · openai.com
How the answer changed A flip happens when any question we’re watching about what AI may soon be able to do changes status: from “Not yet” to “Sort of”, or from “Sort of” to “Yes”.
Sep 7Sort of → Yes
OpenAI says its deployed research agent completes well-defined assignments that would take a skilled person several days and is now used extensively across its research organization.
OpenAI says its deployed research agent completes well-defined assignments that would take a skilled person several days and is now used extensively across its research organization.
Aug 12Not yet → Sort of
Grok Bot’s launched beta runs persistently on cloud computers and completes multi-step workplace assignments, though full-day accuracy remains unproven.
Grok Bot’s launched beta runs persistently on cloud computers and completes multi-step workplace assignments, though full-day accuracy remains unproven.
EarlierOn the Board since launch, answered “Not yet.”