In the popular imagination, the daily life of an artificial intelligence engineer is often visualized as a scene from The Jetsons: a frictionless existence where algorithms anticipate needs, draft correspondence, and manage logistics with superhuman efficiency. The assumption is that those closest to the code would be the first to automate their lives, delegating every mundane task to the digital intelligences they help create. However, a closer look at the habits of industry insiders reveals a startling paradox: the architects of the AI revolution are often the most resistant to letting bots handle the basics.

This hesitation is not born of technophobia, but rather a profound understanding of the technology’s limitations and a desire to preserve cognitive agency. According to a recent report by the Wall Street Journal, professionals deeply immersed in the development of machine learning are increasingly adopting what might be described as an “analog-first” approach to their personal workflows. From handwriting code logic on physical whiteboards to drafting emails character by character, these experts are drawing a sharp line between high-leverage automation and the cognitive drudgery that actually keeps the human mind sharp.

The Trust Deficit: Why Engineers Don’t Auto-Draft

Stella Dong, a machine-learning engineer and co-founder of Reinsurance Analytics, represents a growing cohort of tech workers who refuse to cede their voice to a Large Language Model (LLM). Despite building systems designed to process vast amounts of data, Dong taps out her emails manually. “I don’t trust AI to draft by itself,” Dong told the Wall Street Journal. Her skepticism highlights a critical distinction often lost on the general public: the difference between generating text and communicating intent. While an LLM can predict the next statistically probable word, it cannot replicate the specific interpersonal nuance required in high-stakes business communication.

This reluctance to automate communication is mirrored across the sector. For insiders, the “black box” nature of generative AI—where outputs can be hallucinated or tonally flat—poses a risk that outweighs the time saved. Dong admits to using tools like Copilot for revision, treating the AI as an editor rather than a writer. This reverses the typical consumer behavior, where users prompt an AI to create a draft and then lightly edit it. By writing the draft herself, Dong ensures that the core logic and emotional intelligence of the message remain human, using the AI only to polish the syntax.

The Cognitive Cost of Frictionless Scheduling

Beyond communication, the resistance extends to time management. While AI-driven calendar assistants promise to optimize schedules dynamically, many power users find that manual entry aids memory retention. Dong eschews AI calendar managers, preferring to input appointments herself. This deliberate friction forces a mental acknowledgement of the commitment, making her less likely to forget it. Relying on digital notifications to “shepherd” one through the day can lead to a state of passive compliance, where the worker loses the mental map of their own week.

This phenomenon aligns with broader cognitive science research suggesting that “offloading” mental tasks to digital tools can atrophy the brain’s executive functions. When an AI optimizes a schedule, the human user becomes a passenger in their own workday. By handling the logistics manually, insiders like Dong maintain a sense of control and temporal awareness that automated pings cannot provide. Even in high-tech environments, the most efficient method of booking a meeting often remains a simple phone call—a tactic Dong utilized recently, instructing her business partner to arrange an interview via a human-to-human connection rather than a calendar link.

The Automation Paradox: Just Because We Can, Doesn’t Mean We Should

The tension between capability and utility is quantified in a new report from the McKinsey Global Institute. The firm estimates that existing technology could technically perform 57% of the work hours currently logged by Americans. However, Lareina Yee, a senior partner at McKinsey, warns that this is a “striking but easily misunderstood figure.” The statistic measures technical feasibility, not strategic desirability. We are entering an era where there will be very little that AI cannot do, forcing a philosophical shift toward deciding what it should not do.

“As we redesign work and jobs, you might actually choose not to maximize how much an AI agent or robot does,” Yee notes. This insight suggests a future where companies might deliberately hold back automation to preserve “human in the loop” training. If junior employees no longer draft mundane emails or summarize meetings, they may never develop the foundational skills required to critique the output of an AI assistant later in their careers. The industry is beginning to recognize that “drudgery” often disguises essential learning opportunities.

The Return of Pen and Paper in a Digital World

Perhaps the most counterintuitive trend among AI workers is the resurgence of handwriting. Ziyi Liu, an AI research intern at Microsoft—a company aggressively integrating AI into its Office suite—takes typewritten notes during meetings, ignoring the software’s ability to transcribe and summarize the conversation automatically. For Liu, the act of note-taking is not about preserving a record, but about structuring her own thoughts. “I don’t want to look at a transcript; I just want to do it myself. It makes me feel like I’m in control,” she told the Wall Street Journal.

Ryan Bearden, a marketing consultant who trains businesses on AI tools, takes this a step further by using a Moleskine notebook. In an environment saturated with screens, the physical act of writing signals undivided attention to colleagues. We have all experienced the suspicion that a person typing on a laptop during a meeting is actually multitasking or shopping; a pen and notebook remove that ambiguity. Bearden’s workflow involves sketching physical storyboards before he ever opens a digital tool, using analog methods to solidify his thinking before allowing AI to refine it.

Strategic Friction as a Competitive Advantage

Bearden’s approach underscores a critical “insider” philosophy: AI is a tool for execution, not conception. “AI is a very powerful tool—it’s a hammer and that doesn’t mean everything is a nail,” Bearden says. By forcing the initial creative process to happen on paper, these professionals introduce “strategic friction.” This friction slows down the process just enough to allow for deep work and critical thinking, preventing the rush to a generic, AI-generated solution that might be technically correct but creatively bankrupt.

This methodology reveals a sophisticated understanding of the “knowledge collapse” risk. If every output is generated by an AI trained on the average of human knowledge, the result is inevitably average. To achieve excellence, human intervention must occur before the prompt is even written. The insiders are not rejecting AI; they are disciplining it. They are refusing to let the ease of automation erode the quality of their thought process, choosing to work “slower on purpose” to maintain a competitive edge in creativity and strategy.

The Future of Work: Hybridizing Analog and Digital

As the McKinsey report suggests, the future of work may not be a linear march toward total automation, but a hybrid model where human cognition is treated as a premium asset. We may see a bifurcation in the workforce: those who allow AI to drive their workflows, becoming passive operators of a system they don’t understand, and those who—like Dong, Liu, and Bearden—selectively deploy AI while jealously guarding the manual tasks that keep their minds agile. The irony is palpable: to truly master artificial intelligence, one must first master the discipline of ignoring it.

Ultimately, the habits of these AI workers serve as a canary in the coal mine for the broader economy. They demonstrate that efficiency is not the only metric that matters. Trust, memory, social signaling, and cognitive clarity are often sacrificed at the altar of speed. As AI tools become ubiquitous, the true “power users” will likely be those who know exactly when to put the bot away and pick up a pen.