OpenAI's Greg Brockman: AI Shifting from Chat to Autonomous Task Execution

Gate News message, April 27 — Greg Brockman, president and co-founder of OpenAI, says the next wave of artificial intelligence will move users from chatting with AI bots to assigning real-world tasks. This shift requires enterprises to rethink operational workflows and establish new protocols for security, management, and costs.

Brockman argues that AI must move beyond standalone chat interfaces and integrate directly into existing corporate software to solve problems independently. Recent models have crossed a threshold of usefulness, improving at creating presentations, spreadsheets, and browser-based tasks with minimal instruction. However, granting AI this operational freedom introduces new risks—mistakes could escalate from sending incorrect emails to unauthorized database modifications. To manage this, Brockman explains that employees will transition from executing tasks to overseeing fleets of AI agents, remaining accountable while delegating operational details.

The integration challenge extends to how users interact with AI systems. Models are becoming more intuitive, actively inferring user goals based on context rather than requiring step-by-step instructions. Brockman emphasizes that the competitive moat lies not in individual AI models but in the integrated system itself—comparable to building a car where a superior engine matters little if the rest of the vehicle lacks quality. OpenAI is investing in internal infrastructure and developer tooling to maintain this advantage.

Compute capacity and costs present another critical constraint. As autonomous agents scale, they demand substantial server resources, creating tension between falling compute costs and rising demand. Brockman notes that OpenAI has reduced prices year-over-year while maintaining positive margins, but warns of an approaching “world of compute scarcity” as heavy agent usage approaches rate limits. Enterprise-grade oversight is equally vital; IT departments must maintain full visibility into all AI agents deployed within organizations, managing security, safety, and observability across hundreds of thousands of deployments.

Counterbalancing these optimistic projections, Gartner predicted in June 2025 that over 40% of agentic AI projects will be scrapped by end of 2027 due to rising costs and unclear business value, with only 15% of daily work decisions expected to be made autonomously by 2028. Additionally, quality-adjusted AI model prices have fallen 80% over the past two years as smaller, more efficient models emerged and competition intensified, suggesting workflow integration and governance may become more defensible than raw model capability alone.