Artificial Intelligence

AI Agents Move From Asking to Doing

New OpenAI reports reveal enterprises are shifting AI from information retrieval to task execution, widening the gap between leading 'frontier' firms and others.

StartupHub.ai Staff

By StartupHub.ai Staff, reviewed by Daniel Singer, Editor

4 min read

Graph showing the widening gap in AI output tokens per user between frontier and typical firms over time.
Credit: OpenAI News
Visual TL;DR
OpenAI ReportsContext
new findings detail enterprise AI adoption and increasing complexity
From the article 7 mentionsThis shift, detailed in two new OpenAI reports, highlights a growing divergence between companies that are deeply integrating AI and those that are not.
AI: Asking to DoingDriver
enterprises shifting AI from information retrieval to complex task execution
From the articleInstead of simply asking an AI to draft a presentation, a user can now prompt an agent to gather information from disparate sources, draft the content, and prepare it for review.
Agentic AI RisesCore
From the article 2 mentionsThe core of this evolution lies in the adoption of agentic AI, which allows AI systems to interact with tools, company data, and execute multi-step workflows.
Frontier Firms LeadContext
top 10% of AI users, generating 8.3x more output tokens per user
From the article 3 mentionsAccording to OpenAI's findings, the gap between leading AI adopters, termed 'frontier firms,' and typical organizations is widening.
Beyond EngineeringEffect
agentic work spreads beyond traditional tech roles in enterprises
From the article 3 mentionsWhile software engineering was an early adopter of agentic AI, its adoption is rapidly expanding across knowledge-work functions.
Widening GapOutcome
From the article 2 mentionsAccording to OpenAI's findings, the gap between leading AI adopters, termed 'frontier firms,' and typical organizations is widening.
Contents(4)

Enterprises are accelerating their adoption of artificial intelligence, moving beyond simple question-answering to delegating complex tasks. This shift, detailed in two new OpenAI reports, highlights a growing divergence between companies that are deeply integrating AI and those that are not. The core of this evolution lies in the adoption of agentic AI, which allows AI systems to interact with tools, company data, and execute multi-step workflows.

According to OpenAI's findings, the gap between leading AI adopters, termed 'frontier firms,' and typical organizations is widening. Frontier firms, representing the top 10% of monthly AI users, now generate 8.3 times more output tokens per active user compared to average firms. This metric serves as a proxy for the depth and complexity of AI engagement, indicating that leading companies are not just using AI more frequently but are entrusting it with more substantial work.

The Rise of Agentic AI

The transition from AI as an assistant to AI as an executor is powered by advancements like OpenAI's Codex and ChatGPT Enterprise. These tools can now access company-specific context and integrate with existing business tools, such as CRMs or internal playbooks. Instead of simply asking an AI to draft a presentation, a user can now prompt an agent to gather information from disparate sources, draft the content, and prepare it for review. As of June, Codex accounted for 64% of combined Codex and ChatGPT output tokens among enterprise customers, signaling a significant move towards delegated, multi-step tasks.

The Widening Frontier Gap

The disparity in AI adoption is stark. Frontier firms are significantly more likely to utilize advanced AI capabilities. For instance, 21% of active users at these leading firms use Plugins weekly, compared to just 9% at typical firms. This suggests that companies at the forefront are actively integrating AI with external tools and internal systems to automate complex processes. StartupHub.ai data shows that while OpenAI itself scores a strong 84/100 in our proprietary index, the widespread adoption of specialized tools like Zapier (score 77/100) and UiPath (score 85/100) indicates a broader market trend towards AI-powered automation solutions.

This gap isn't confined to tech giants. The reports indicate that intensive AI use is prevalent across various industries and company sizes, not just within technology sectors. Companies that are more aggressive AI adopters also tend to exhibit stronger financial performance, holding more assets, employing more people, and investing more heavily in research and development, according to an analysis of U.S. public companies in the accompanying working paper.

Agentic Work Spreads Beyond Engineering

While software engineering was an early adopter of agentic AI, its adoption is rapidly expanding across knowledge-work functions. Since February, weekly active enterprise Codex users have seen explosive growth in fields like legal (108x increase), sales (41x increase), and recruiting (41x increase). Marketing departments also saw a 26x surge in usage. This broadens the impact of AI beyond technical roles, enabling professionals in diverse fields to automate tasks and accelerate project timelines.

Early-career employees are leading this charge. Contrary to some surveys that suggest higher AI usage among executives, OpenAI's data indicates that junior employees send more AI-generated messages per week. This suggests that newer entrants to the workforce may have a natural aptitude for integrating AI into their workflows, presenting an opportunity for leaders to identify and disseminate these effective practices across all organizational levels.

Bridging the Gap

Leaders aiming to narrow this frontier gap must focus on more than just providing access to AI models. The reports suggest a practical agenda: connecting AI agents to relevant company context and tools, establishing clear governance and permission structures, and facilitating the transformation of individual effective workflows into shared organizational practices. By enabling agents to securely access company data and execute tasks with appropriate oversight, enterprises can fully transition from AI as an information source to AI as a core operational engine.

Mentioned in this story

Markets

UiPathNYSE:PATH

Sources

TopicsOpenAIChatGPTCodexPluginsZapierUiPath

StartupHub.ai StaffDaniel SingerWritten by StartupHub.ai Staff, Editorial teamReviewed by Daniel Singer, EditorAbout

StartupHub.ai Staff, Editorial team

The staff writers of StartupHub.ai, ranging from investment analysts to avid AI tool users, early adopters and critical enthusiasts. Backgrounds span engineering, business and the arts. We hold every piece to rigorous standards of research and review.

Daniel Singer, Editor

Daniel Singer is an AI expert who has been writing about, covering and building AI technology since 2018.

More from Daniel Singer

Get the free daily brief every morning

Every morning the day's AI news, new AI startups and new funding rounds. Free and takes a few minutes to read, unsubscribe at any time.

Join 64,000 other subscribers.

Test your Agent Readiness now

Paste your website. How well ChatGPT, Claude and the other agents can read and use a site, in about a minute.

The videos behind the news, one flip at a time

The feed mixes the videos from our articles with startup demos and you flip through them like TV channels for free with no signup.