OpenAI CEO Sam Altman has acknowledged that his earlier predictions about the speed at which artificial intelligence would transform businesses and the economy were too ambitious. In a recent podcast conversation with David Senra, Altman reflected on the period following the launch of GPT 4 in 2023 and said he had expected much more disruption in the software industry much sooner than what actually occurred.
Altman said he had anticipated that software companies and established business models would become vulnerable to rapid disruption almost immediately after GPT 4 became available. Instead, the adoption of AI across businesses has taken longer, with companies continuing to rely on existing systems, processes and working habits.
The comments highlight a distinction between technological progress and economic adoption. AI models have improved substantially over the past several years, but organisations do not automatically change their operations as soon as a new technology becomes available. Companies need to evaluate costs, security, reliability, staff capabilities and integration with existing systems before adopting new tools on a large scale.
Altman described this slower transition as a result of economic inertia. People and organisations often prefer familiar tools and established workflows, even when newer technologies offer significant capabilities. This inertia can delay the broader economic impact of technological breakthroughs.
The acknowledgement represents a notable adjustment from some of the earlier predictions surrounding generative AI. After the release of GPT 4, expectations grew that AI would rapidly transform programming, software development, customer service, research and other knowledge based activities. While those changes are underway, the pace has varied significantly between industries and companies.
OpenAI itself has been examining this changing pattern of AI adoption. In a June 2026 report on its Codex coding agent, the company said the use of AI agents inside OpenAI had expanded rapidly across departments. By that point, non technical teams such as legal, recruiting and finance were also using Codex for tasks beyond traditional software development.
The development of AI coding tools provides another example of how quickly the technology landscape can change. Altman recalled that OpenAI was initially behind Anthropic's Claude Code in the coding agent market. He described OpenAI's decision to build and compete in that area despite the existing lead held by Claude Code as an unusually high risk strategy. Reports summarising his comments say he described the effort as a "kamikaze mission."
Claude Code has become an important example of the transition from AI that simply generates text to AI systems that can carry out longer and more complex tasks. Anthropic reported that by May 2026, more than 80 percent of the code merged into its own codebase had been authored by Claude. The company said this represented a major shift from the low single digit share recorded before Claude Code's research preview.
OpenAI has reported a similar internal transformation. The company said Codex use grew sharply among developers and non developers during 2026, with employees increasingly using the system for tasks that could take humans more than an hour to complete. This suggests that while broader economic adoption may be slower than Altman originally expected, adoption within technologically advanced organisations can still accelerate rapidly once the right tools are available.
Altman's revised view does not suggest that AI will have a smaller long term impact. Rather, his comments indicate that the timing of that impact may extend over a longer period. The technology can continue improving rapidly while businesses take more time to reorganise around it.
This distinction is important for companies making investment decisions. AI adoption often requires more than purchasing access to a chatbot. Businesses may need to redesign workflows, train employees, change software infrastructure, establish security controls and determine how human decision making will interact with AI systems.
Recent enterprise research also shows that AI adoption is spreading, but unevenly. OpenAI said in August 2026 that enterprises are increasingly moving from simple AI assistance toward systems that can execute longer tasks. The company also reported that advanced organisations are adopting agentic AI more rapidly than less advanced firms.
The economic impact of AI may therefore arrive in stages rather than through one sudden transformation. Early adoption can begin with individual productivity tools before moving into company wide automation, AI agents and redesigned business processes.
For workers, this gradual transition could provide additional time to adapt, although it does not eliminate the need for reskilling. Routine knowledge work may increasingly be assisted or automated, while demand grows for employees who can manage AI systems, evaluate outputs and combine technical expertise with domain knowledge.
Altman's comments also underline the uncertainty involved in predicting the future of rapidly evolving technology. Even executives directly involved in developing advanced AI systems can misjudge how quickly society and businesses will change.
The AI industry itself continues to move at a much faster pace than many traditional businesses. New models, coding agents and other AI systems are being introduced quickly, while companies outside the technology sector often take longer to evaluate their practical value.
For OpenAI, the challenge is therefore twofold. The company needs to keep improving its models and products while also ensuring that businesses can adopt them effectively. Competition from Anthropic, Google and other AI companies makes that task increasingly important.
Altman's admission provides a more measured view of the AI transition. The technology may continue advancing rapidly, but the economic transformation associated with it could take longer than early predictions suggested.
The main lesson from his comments is that technological capability and adoption are not the same thing. AI can become increasingly powerful while businesses, workers and consumers take years to change their habits and processes.
That slower adoption could mean a more gradual transition for the global economy. It may also give businesses additional time to build appropriate systems, train employees and determine where AI provides genuine value.
At the same time, the rapid progress of coding agents such as Codex and Claude Code shows that adoption can accelerate quickly once organisations discover clear and measurable benefits. The gap between experimentation and widespread operational use could therefore narrow considerably in specific industries.
Sam Altman's revised timeline is consequently less a retreat from the importance of AI than a recognition that economic change follows its own pace. The coming years will show whether adoption continues gradually or accelerates as AI becomes more reliable, affordable and deeply integrated into everyday business operations.