By Lina Irawan 
The term "friction" itself is a deliberate choice, acknowledging the inherent resistance that develops within any rapidly scaling organization. Unlike traditional corporate structures that often obscure or ignore bureaucratic hurdles, OpenAI has chosen to confront them head-on, giving them a name and a dedicated resolution pathway. This system has the power to supersede almost any ongoing team activity, often carrying the implicit or explicit authority of CEO Sam Altman or President Greg Brockman, ensuring that reported issues receive immediate and high-priority attention.
The process is deceptively simple in its initiation: an employee emails [email protected] to report an internal bottleneck. However, the scope of issues deemed worthy of this channel is broad and deeply insightful into the challenges of an AI powerhouse. These reports can range from critical technical systems malfunctioning, like a core AI model training pipeline experiencing unexpected delays or a crucial data processing cluster failing, to fundamental process breakdowns, such as a convoluted approval flow for deploying new features that hinders agile development. Beyond the purely technical, it also encompasses more mundane but equally impactful office-related frustrations, like a shortage of essential IT vending machines or inadequate parking solutions, which, if left unaddressed, can chip away at employee morale and productivity.
Once an email lands in the friction@ inbox, it doesn’t disappear into a void. Instead, it is triaged directly by OpenAI’s leadership team. This hands-on approach from the very top signals the company’s profound commitment to eradicating inefficiencies. Leadership reviews the submissions, assessing their potential impact on the company’s core mission and operational velocity. Issues deemed significant enough are then escalated, often directly involving Altman or Brockman to ensure swift and decisive resolution. For instance, reports of perpetually full parking lots, a seemingly minor logistical issue, prompted a pilot program to prioritize spots for employees with longer commutes, demonstrating a practical response to everyday employee concerns. More critically, when employees highlighted a cumbersome process for reliably granting API credits at scale—a vital component for external developers and partners—the "friction" system spurred a re-evaluation and streamlining of that entire workflow, directly impacting OpenAI’s ecosystem growth and developer experience.
A former OpenAI employee, speaking to Fortune, lauded the "friction" system as an exceptionally effective method for leadership to gain unfiltered, ground-level feedback. This direct conduit allows issues to be identified and addressed before they can metastasize into larger, more intractable problems, thereby maintaining the company’s momentum. This proactive problem-solving mechanism is increasingly vital as OpenAI undergoes explosive growth, with its headcount projected to swell to over 8,000 employees by the end of this year, spanning offices across the U.S. and globally. Scaling at such a pace often introduces layers of bureaucracy and communication breakdowns, making a direct feedback loop indispensable.
"It’s a good thing—literally anybody can complain about big company bullsh*t directly to Sam and they can act on it," the former employee stated, underscoring the system’s democratic and empowering nature. "If you want to ruthlessly cut through bureaucracy, you have to be ruthless about it." This sentiment reflects a core tenet of OpenAI’s culture: an aggressive intolerance for anything that impedes progress, especially in a field where speed to market and continuous innovation are paramount. This "ruthless" approach is seen as a necessary evil to maintain the agile, startup-like velocity even as the company rapidly matures into a global enterprise.
While the "friction" email address had existed within OpenAI for some time, its true potential wasn’t fully realized until the fall of 2025. It was then that Fidji Simo, a seasoned tech executive with a distinguished background at Facebook/Meta, joined OpenAI as CEO of Applications. Simo, known for her operational prowess and user-centric approach, began her tenure with a comprehensive three-month "listening tour." During this period, she engaged deeply with employees across various departments to understand the operational pain points and systemic challenges within the rapidly expanding organization. Consistently, a central theme emerged from these conversations: concerns about maintaining agility and moving quickly as the company scaled.
Simo’s insights into organizational decay resonated deeply with the challenges OpenAI faced. "Companies rarely become bureaucratic all of a sudden," Simo told Fortune. "It happens one unnecessary meeting, one extra approval, one small frustration at a time. Each makes it only 1 or 2% harder to do your job, which is why they’re easy to ignore, but those frictions compound." This understanding highlighted the insidious nature of minor inefficiencies, which, individually negligible, collectively create significant drag on productivity and innovation. Her experience at large tech companies likely provided her with a keen eye for these subtle, yet destructive, cumulative effects.
Recognizing the latent power of the friction@ address, Simo moved to institutionalize and optimize the process. She tasked Irina Kofman, OpenAI’s VP of strategic initiatives and operations, with the critical responsibility of monitoring the inbox, triaging incoming emails, and ensuring that select issues were not only addressed but demonstrably resolved. Crucially, Simo also initiated a monthly, company-wide Slack update, transparently tracking the progress and resolution of "friction" items. This regular communication served a dual purpose: it assured employees that their concerns were being taken seriously and acted upon, fostering trust and engagement, and it also held leadership accountable for follow-through. Although Simo departed OpenAI in early July to focus on her health and her healthcare startup, ChronicleBio, the formalized friction@ process and the company-wide Slack updates remain, a testament to the enduring value and impact of her strategic interventions.
While the concerted effort to root out inefficiencies through the "friction" system has undeniable benefits and has become an integral part of OpenAI’s culture, it is not without its detractors. The intensity and urgency it often demands can create significant stress, highlighting the delicate balance between speed and sustainable operations.
One former team leader at OpenAI, who claimed to have been on the receiving end of "dozens" of "friction" emails over their several years at the company, articulated a different perspective. They argued that the process, while well-intentioned, was often so disruptive that it inadvertently created new inefficiencies. "The individual who reports the issue probably thinks it’s the single most important thing, but to teams like mine that are constantly underwater, under-resourced, and over-leveraged, it’s the single most disruptive thing you could do," the former employee explained. This highlights a fundamental tension: what one person perceives as a critical bottleneck, another sees as a sudden, unplanned priority shift that derails existing workstreams and strains already stretched resources.
OpenAI is renowned for its intense work culture, where working long hours—often extending well past dinnertime into the night—is a common expectation. In such an environment, "friction" emails can trigger unwelcome "fire drills," demanding immediate attention and diverting focus from ongoing projects. "If Sam and Greg think something is worthy, if it’s defensible, then essentially a ‘friction’ gets started, and the party who is responsible for the block is then more or less just told to shut up and do it," the person elaborated. "It’s no joke, a drop everything kind of deal." This description paints a picture of a command-and-control dynamic, where top-down directives supersede departmental autonomy, potentially leading to resentment and burnout among managers and teams. The implication is that while the problem gets solved, the manner of resolution can be jarring and taxing on those tasked with executing it.
However, the prevailing sentiment within OpenAI and the broader tech industry dictates that the expediency of shaving off every possible drag on speed trumps these concerns. The competition in the race to build and release the latest, most advanced AI models is extraordinarily fierce, creating an existential imperative for rapid iteration and deployment. In this high-stakes environment, "fire drills" and company-wide "lockdowns" are often accepted as an intrinsic part of the job. For example, in December 2025, OpenAI famously instituted an internal "code red," a dramatic call to arms that pushed employees to double down on strategic product initiatives. This aggressive maneuver was explicitly aimed at blunting the advances of formidable rivals like Google and Anthropic, underscoring the absolute priority of speed and competitive edge.
OpenAI’s "friction" emails also draw a compelling parallel to a legendary, and often dreaded, practice at another tech giant: the infamous question mark emails from Amazon founder Jeff Bezos. When an Amazon customer emailed Bezos directly at [email protected] with a complaint or issue, if Bezos deemed the matter significant, he would forward the email to the responsible team with a single, ominous question mark. This was internally known as a "Jeff B escalation," and it signaled an urgent, non-negotiable directive. The receiving team was expected to prioritize a deep dive into the issue, provide a comprehensive explanation, and implement immediate corrective action. The "Jeff B escalation" was a powerful tool for maintaining Amazon’s relentless customer obsession and holding teams accountable.
In this tradition, OpenAI’s friction@ email address serves a similar function, albeit with an internal focus rather than an external, customer-facing one. Both systems leverage the personal authority of the company’s highest leaders to cut through organizational inertia, demanding immediate attention and resolution for perceived blockages. While Bezos’s system aimed at perfecting the customer experience, OpenAI’s targets internal operational bottlenecks that could slow down its mission to develop advanced AI. Both practices highlight a common thread in ambitious tech companies: a willingness to bypass traditional hierarchies and empower top leadership to directly intervene in day-to-day operations to achieve strategic imperatives. This approach, while potentially disruptive, is seen as a necessary evil to maintain a competitive edge and prevent the sclerosis that often afflicts large organizations. As OpenAI continues its meteoric rise, the "friction" system will undoubtedly remain a defining, if sometimes polarizing, feature of its quest for AI supremacy.
