In what might be the most telling anecdote about the gap between artificial intelligence’s promise and its present-day utility, OpenAI’s own board chairman, Bret Taylor, recently admitted that his preparation materials for board meetings at the world’s most prominent AI company were written entirely without the help of AI. The revelation, first reported by Business Insider, has sparked a broader conversation among technology executives and industry watchers about whether the people building AI tools actually trust them enough to use them in high-stakes settings.
Taylor, who became OpenAI’s board chairman in the aftermath of the dramatic boardroom coup that briefly ousted CEO Sam Altman in November 2023, is no stranger to enterprise technology. He previously served as co-CEO of Salesforce and was a key figure in the development of Google Maps. His credentials as a technologist are beyond question. Yet when it came to the serious work of governing a company valued at more than $300 billion, he reached for the same tools that executives have relied on for decades: traditional word processors, spreadsheets, and human-drafted memos.
The Cobbler’s Children Have No Shoes — Or Do They?
The irony is hard to miss. OpenAI has spent the past several years positioning its products — ChatGPT, GPT-4, and a growing roster of enterprise-grade tools — as indispensable aids for knowledge workers. The company’s pitch to corporate America has centered on the idea that AI can draft documents, summarize complex information, and accelerate decision-making at every level of an organization. Board governance, with its dense legal filings, financial summaries, and strategic analyses, would seem like a natural fit for AI assistance.
Yet Taylor’s approach suggests that even at the highest levels of the AI industry, there remains a trust deficit when the stakes are sufficiently high. Board meeting preparation at a company like OpenAI involves sensitive financial data, confidential strategic discussions, and legal considerations that carry real consequences if errors slip through. According to Business Insider, the materials Taylor worked with were prepared through conventional means, without AI-generated drafts or summaries playing a meaningful role in the process.
A Pattern That Extends Well Beyond OpenAI
Taylor’s case is not an isolated one. Across Silicon Valley and Wall Street alike, there is a growing body of evidence that senior executives and board members remain cautious about incorporating AI into their most consequential workflows. A 2025 survey by McKinsey found that while more than 70% of large enterprises had deployed generative AI tools in some capacity, usage dropped sharply at the C-suite and board level, where concerns about accuracy, confidentiality, and liability loom largest.
The reasons are not difficult to understand. Large language models, for all their impressive capabilities, are still prone to hallucinations — generating plausible-sounding but factually incorrect information. In a board setting, where directors have fiduciary duties and where a single misstatement in a financial summary could have legal ramifications, the tolerance for error is essentially zero. “You can’t have a model that’s right 95% of the time when you need it to be right 100% of the time,” one corporate governance attorney told colleagues at a recent conference. The sentiment captures a widespread feeling among the people who sit in boardrooms.
Bret Taylor’s Dual Role Complicates the Narrative
What makes Taylor’s situation particularly interesting is that he is not merely the chairman of OpenAI’s board. He is also the CEO and co-founder of Sierra AI, a startup focused on building AI agents for enterprise customer service. Sierra has raised hundreds of millions of dollars and counts major brands among its clients. Taylor is, in other words, simultaneously one of the most prominent evangelists for AI in the enterprise and someone who, in his own governance work, opted not to use it.
This duality raises questions that go beyond personal preference. If the chairman of OpenAI — a man who has spent years building and selling AI products — does not rely on AI for his own high-stakes document preparation, what message does that send to the Fortune 500 CIOs and CTOs who are being asked to deploy these tools across their organizations? The answer, according to several industry analysts, is more nuanced than it might first appear.
The Distinction Between ‘Not Ready’ and ‘Not Appropriate’
Defenders of Taylor’s approach argue that choosing not to use AI for board preparation is not the same as saying AI is not ready for serious work. Board governance sits at the intersection of legal, financial, and strategic concerns where human judgment, institutional knowledge, and accountability are paramount. AI tools today are excellent at first-draft generation, brainstorming, and processing large volumes of information, but they are not yet capable of exercising the kind of judgment that board-level work demands.
“There’s a difference between using AI to draft a marketing email and using it to prepare materials that will inform decisions worth billions of dollars,” said one venture capitalist who has served on multiple technology company boards. “The risk-reward calculus is completely different.” This perspective suggests that the issue is less about AI’s shortcomings and more about the specific requirements of corporate governance — a domain where precision, confidentiality, and legal defensibility are non-negotiable.
OpenAI’s Internal Culture and the Pressure to Practice What You Preach
Inside OpenAI, the revelation has reportedly generated some internal discussion, according to people familiar with the company’s culture. OpenAI has long encouraged its employees to use its own products extensively — a practice known in the technology industry as “dogfooding.” The company’s researchers and engineers routinely use ChatGPT and internal tools for coding, writing, and analysis. The expectation that AI should be woven into daily workflows is deeply embedded in the organization’s identity.
But board-level governance occupies a different category. OpenAI’s board, reconstituted after the November 2023 crisis, includes members with backgrounds in technology, national security, and academia. The board’s responsibilities have only grown more complex as OpenAI has transitioned from a nonprofit to a for-profit structure, raised tens of billions of dollars, and faced increasing regulatory scrutiny from governments around the world. In this context, the decision to prepare materials without AI assistance may reflect not a lack of confidence in the technology but rather a sober assessment of what the moment requires.
What This Means for the Broader Enterprise AI Market
The implications of Taylor’s approach extend well beyond OpenAI’s boardroom. Enterprise AI adoption has been one of the defining business stories of the past three years, with companies spending billions on licenses, infrastructure, and integration. Yet the question of where AI is appropriate — and where it is not — remains largely unresolved. Taylor’s example provides a data point that is difficult for the industry to ignore: even the people closest to the technology recognize that there are domains where human-only workflows remain preferable.
This does not mean that AI will never be used in board governance. Several startups, including Diligent and BoardVantage, are already developing AI-powered tools specifically designed for board management, with features like automated minute-taking, document summarization, and compliance monitoring. The market for these tools is expected to grow significantly in the coming years. But the adoption curve at the board level is likely to be slower and more cautious than in other parts of the enterprise, precisely because the consequences of failure are so high.
The Honest Signal in Taylor’s Choice
Perhaps the most valuable thing about Taylor’s admission is its honesty. In an industry where executives routinely overstate the capabilities of their products and where the pressure to appear AI-forward is intense, a board chairman acknowledging that he prepared for meetings the old-fashioned way is a refreshing dose of candor. It suggests that the leaders of the AI industry are capable of distinguishing between what their technology can do today and what it should be trusted to do.
For enterprise buyers weighing their own AI strategies, the lesson may be straightforward: AI is a powerful tool, but it is not yet a universal one. The most sophisticated users of the technology — including the people who build it — still exercise judgment about when to deploy it and when to rely on human expertise alone. That kind of discernment, rather than blanket adoption or blanket skepticism, may be the most mature approach to a technology that is still finding its place in the highest-stakes corners of the business world.