<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/ai-agents-are-creating-more-work-not-less-and-openai-s-own-numbers-back-it-up-y9f08ulpn" -->

---
title: AI agents are creating more work, not less — and...
description: OpenAI's internal data shows research agents logging 3.1 agent-workdays per human workday, with median researchers spending over $600/day on inference (90th...
canonical: https://daily.dev/posts/ai-agents-are-creating-more-work-not-less-and-openai-s-own-numbers-back-it-up-y9f08ulpn
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: AI agents are creating more work, not less — and OpenAI’s own numbers back it up | daily.dev
og:description: OpenAI's internal data shows research agents logging 3.1 agent-workdays per human workday, with median researchers spending over $600/day on inference (90th...
og:url: https://daily.dev/posts/ai-agents-are-creating-more-work-not-less-and-openai-s-own-numbers-back-it-up-y9f08ulpn
og:image: https://api.daily.dev/og/posts/Y9F08uLpn.png
og:image:alt: AI agents are creating more work, not less — and OpenAI’s own numbers back it up
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# AI agents are creating more work, not less — and OpenAI’s own numbers back it up

**[The New Stack](https://daily.dev/sources/newstack)** · 5 min read · 0 upvotes · 0 comments

## Summary

OpenAI's internal data shows research agents logging 3.1 agent-workdays per human workday, with median researchers spending over $600/day on inference (90th percentile over $7,000). Despite rising agent activity across all six research work categories since January, humans still had to intervene on more than half of successful tasks taking four to eight hours, and supervision — not raw agent output — is becoming the real bottleneck. Security concerns around the Astra model triggered access restrictions in July and August after outages and a near-'Critical' cybersecurity threshold finding, causing a 59.2% drop in Astra GPU allocation that was quickly absorbed by other models, showing how fluidly workloads shift when one system component is locked down.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://thenewstack.io/openai-agent-research-bottleneck>

## Questions this post answers

### Why did OpenAI restrict access to its Astra model in August?

OpenAI tightened access to Astra on August 7 after early evidence suggested the model could reach the 'Critical' cybersecurity threshold defined in its Preparedness Framework. The restrictions confined Astra to higher-security research areas and added safeguards, which developers may notice as unexpected API interruptions. This followed a July 20 outage that had already forced OpenAI to take its training container service offline temporarily.

_daily.dev helps developers track API-affecting safety restrictions like these before they cause unexpected outages._

### How much does it cost OpenAI researchers to run AI agents on inference at API prices?

The median researcher at OpenAI spends more than $600 per day on inference at API prices, while researchers in the 90th percentile spend more than $7,000 per day. This spending accompanies a rise in agent-workday usage, which reached 3.1 agent-workdays logged for every one human workday across the research organization by mid-August.

_Teams weighing agent-driven inference costs can follow real usage benchmarks like this on daily.dev._

### Do AI coding agents actually reduce the amount of work humans have to do?

Not necessarily. OpenAI found that despite improving success rates between January and July, humans still had to intervene on more than half of successful agent tasks that would have taken a person four to eight hours to complete. Running more agents in parallel increases total work happening at once but also increases the supervisory burden of reviewing code diffs and catching runs that go off track.

_daily.dev keeps builders informed on real-world data about agent supervision overhead before adopting these workflows._

## Similar posts on daily.dev

- [Companies rein in AI spending as running costs mount](https://daily.dev/posts/companies-rein-in-ai-spending-as-running-costs-mount-1dwposyjf) · TechCentral · 0 upvotes · 0 comments
- [Are the Costs of AI Agents Also Rising Exponentially? — Toby Ord](https://daily.dev/posts/are-the-costs-of-ai-agents-also-rising-exponentially-toby-ord-bjsdu9aoh) · Hacker News · 2 upvotes · 0 comments

---

Tags: [#security](https://daily.dev/tags/security), [#ai-agents](https://daily.dev/tags/ai-agents), [#productivity](https://daily.dev/tags/productivity), [#openai](https://daily.dev/tags/openai), [#ai-infrastructure](https://daily.dev/tags/ai-infrastructure)

[View this post on daily.dev](https://daily.dev/posts/ai-agents-are-creating-more-work-not-less-and-openai-s-own-numbers-back-it-up-y9f08ulpn)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"AI agents are creating more work, not less — and OpenAI’s own numbers back it up","url":"https://daily.dev/posts/ai-agents-are-creating-more-work-not-less-and-openai-s-own-numbers-back-it-up-y9f08ulpn","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-agents-are-creating-more-work-not-less-and-openai-s-own-numbers-back-it-up-y9f08ulpn"},"datePublished":"2026-09-07T19:12:57.605Z","dateModified":"2026-10-15T02:22:36.160Z","description":"OpenAI's internal data shows research agents logging 3.1 agent-workdays per human workday, with median researchers spending over $600/day on inference (90th...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/0fe690c071ab64d1b1818acb16f23f87?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/0fe690c071ab64d1b1818acb16f23f87?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The New Stack","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The New Stack","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/newstack","url":"https://daily.dev/sources/newstack"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-agents-are-creating-more-work-not-less-and-openai-s-own-numbers-back-it-up-y9f08ulpn","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"security,ai-agents,productivity,openai,ai-infrastructure","timeRequired":"PT5M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The New Stack","item":"https://daily.dev/sources/newstack"},{"@type":"ListItem","position":3,"name":"AI agents are creating more work, not less — and OpenAI’s own numbers back it up"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-agents-are-creating-more-work-not-less-and-openai-s-own-numbers-back-it-up-y9f08ulpn#faq","mainEntity":[{"@type":"Question","name":"Why did OpenAI restrict access to its Astra model in August?","acceptedAnswer":{"@type":"Answer","text":"OpenAI tightened access to Astra on August 7 after early evidence suggested the model could reach the 'Critical' cybersecurity threshold defined in its Preparedness Framework. The restrictions confined Astra to higher-security research areas and added safeguards, which developers may notice as unexpected API interruptions. This followed a July 20 outage that had already forced OpenAI to take its training container service offline temporarily. daily.dev helps developers track API-affecting safety restrictions like these before they cause unexpected outages."}},{"@type":"Question","name":"How much does it cost OpenAI researchers to run AI agents on inference at API prices?","acceptedAnswer":{"@type":"Answer","text":"The median researcher at OpenAI spends more than $600 per day on inference at API prices, while researchers in the 90th percentile spend more than $7,000 per day. This spending accompanies a rise in agent-workday usage, which reached 3.1 agent-workdays logged for every one human workday across the research organization by mid-August. Teams weighing agent-driven inference costs can follow real usage benchmarks like this on daily.dev."}},{"@type":"Question","name":"Do AI coding agents actually reduce the amount of work humans have to do?","acceptedAnswer":{"@type":"Answer","text":"Not necessarily. OpenAI found that despite improving success rates between January and July, humans still had to intervene on more than half of successful agent tasks that would have taken a person four to eight hours to complete. Running more agents in parallel increases total work happening at once but also increases the supervisory burden of reviewing code diffs and catching runs that go off track. daily.dev keeps builders informed on real-world data about agent supervision overhead before adopting these workflows."}}]}
```

