Skip to main content

Your Best Workers Don’t Need AI—and You Might Lose Them Because of It.

AI is covering up a global digital literacy crisis, and stands to make things worse.

Black and white image of a classic Apple II computer on display in Wrocław, Poland.


By Nathan Upchurch

The most productive staff at any organization have long been automating their work. Maybe they didn’t tell you because they were likely to be punished for their efficiency with extra work. Maybe you just didn’t notice. But while others treat their computers like an obstacle, this small subset of workers recognise them for the miracle tools that they are, and take learning how to use them seriously as a result. These are the people who you’ve relied on for years when you need to get something difficult or novel done. They’ve been knocking out reports, designing processes, building tools, and improving operations with the breadth of their skill with a computer long before LLMs became popular. They’re the pinch-hitters you can always count on, and AI mania from the top may just push them out the door.

These are the employees who don’t need AI to do a simple find and replace, produce charts, find invalid data, or convert documents. Instead of memorizing a series of steps for particular work tasks, they learned—truly learned—to use spreadsheet software, regex, pandoc, and more. They took the promise of computing as a ‘bicycle for the mind’ seriously, and became fiercely competent simply by having the willingness to learn. And why shouldn’t they? After all, in what other industry, trade, or role beside white-collar knowledge work can workers get away with not knowing how to use their tools properly? Regardless, the reason that makes these individuals load-bearing in organizations world-over also makes LLMs seductive to leadership: a digital literacy crisis of terrifying scale.

According to a 2020 report from the National Skills Coalition, nearly one in three US workers lack ‘foundational digital skills’. Further, an international research study tested over 215,000 people throughout 2011–2015 finding that only five percent of the adult population tested could perform “difficult” tasks on a computer. What might a “difficult” task look like? Here’s one example: scheduling a meeting using information contained in several emails. Yikes. Information on computer literacy among executives is thin on the ground, but when you ask their employees the picture is bleak, with over sixty percent calling their bosses “digitally illiterate.”

While 91% of us keep a computer in our pocket, statistically speaking, the average American is more likely to have a serious mental illness than they are to be proficient with a computer. With such a gross skill deficiency spanning from front-line staff to corner offices, it’s only to be expected that elementary knowledge comes to be treated as though it were within the domain of specialists. While you might imagine that this harrowing systemic deficit might be a cause for expanded on-the-job-training or educational initiatives it seems that, instead, unwitting denizens of corner-offices around the world instead see AI as a democratization of this ‘specialist knowledge’, and set out to empower their team to do more with less while tokens are cheap. Under the circumstances, and in the face of the lofty promises of the AI industry, leadership simply isn’t equipped to understand staff pushback. Likewise, why would bosses be at all interested in naysaying from peons while attending pro-AI seminars hosted by armies of highly paid consultants and tech CEOs, and hearing their praises sung by sycophantic chatbots? Only the people on the ground seem to recognize the very real problems encountered when applying AI to operations.

Non-AI tools used by skilled workers for decades allow for exacting control and precision that is relinquished when using AI. LLMs make changes that weren’t asked for, skip ones that were, and the “hallucination” problem introduces subtle errors into just about anything a chatbot touches. Why would someone who knows how to write a quick `VLOOKUP()` formula hand the task to an AI and be liable for its mistakes when they could spend thirty seconds typing and be assured that the results are correct? Why would they risk unwanted changes to a transcript when using the built-in ‘find and replace’ function is faster than typing a prompt? Leadership might push for more to get done faster with AI tools but, even in cases where this is possible, they aren’t the ones whose livelihoods are jeapordized by the resulting output quality. After all, no one is going to be pointing a finger at Claude when the wrong numbers are reported to the board.

Even in an AI-driven Shangri-La where output is perfect, usage itself might cripple an organization’s ability to innovate. Staff who are not naturally inclined to learn simply never do, and skilled staff slowly lose the abilities that once represented a competitive advantage for the organization. It’s easy to imagine organizations beginning to operate at the pleasure of AI companies as a resulting skill hemorrhage reaches its terminal phase.

For workers, another major pain point is that learning and problem-solving are key job benefits. Not only does AI threaten their livelihood with untrustworthy output while robbing them of the skills they worked so hard to obtain, it excises the joy and sense of accomplishment they once found in their work. Surely we could make a business case relating to this final aspect, but should we really have to?

While the c-suite is puzzled by AI-resistance among technically inclined staff, team members are similarly confused as to why higher-ups insist on pushing a technology that they simply don’t need. They’ve learned to do their jobs in the cockpit of a Boeing 747 and are having a Fisher-Price steering wheel shoved into their hands instead—by someone who hasn’t so much as taxied down the runway in twenty years, and is unwilling to entertain the idea that maybe their darling tool of the moment simply isn’t up to the task.

This is perhaps the most psychologically devastating aspect of the current AI frenzy. Staff with skills that their employers do not possess, who do work that their employers do not do, are being implicitly told that they are too stupid to know how best to do their own job. Rather than steering the company, AI-enthusiast bosses world-over seem to be down in the prep-kitchen pressing a Slap Chop into the palm of each cook as if they can’t dice onions on their own. It’s a classic leadership flaw, and the cornerstone of founder’s syndrome: giving skilled workers top-down solutions instead of handing them the problems that they are uniquely positioned and qualified to solve. Just about anyone who has worked for someone else can understand the demoralization that this causes, and imagine the inevitable churn and loss of institutional knowledge that is likely to result.

It seems difficult for leaders to face up to the unfortunate truth that the ongoing AI craze lays bare: that we’re facing a major crisis of skill. Under the circumstances, is it any surprise that 70% of digital transformation efforts fail, or that as 98% of companies experiment with AI, a mere 4% have used it to “generate substantial value?” As AI is being slapped onto products and processes en masse, basic digital process-improvements risk being skipped, which could result in the very problems AI vendors claim that their products can help solve.

The AI industry seems like both a symptom and a snake-oil: an attempt to spackle over the cracks in a foundational lack of computer literacy. Meanwhile, highly independent workers with strong technical ability are left feeling demoralized by the attempt, and leaders risk losing the only people who can not only speak reason to power, but problem-solve without using a single token and continue to innovate when VC dollars dry up alongside aggressive pricing subsidization.

In a world where we’re constantly being told that we’d better get on board or be left behind, bold leadership would tackle this crisis at its root. It would look like extensive computer education, starting by helping employees develop a working mental model of how computers operate and common UI patterns; identifying highly computer literate workers and empowering them to solve difficult problems; doing the hard work of cleaning up data with a human brain; and properly preparing systems and processes for reliable, time-tested automation tools that don’t bill by the token. It’s not too late to win back employee trust and build organizations known for trustworthy data, real efficiency, and human attention where it matters. Bold leadership will slow down while everyone else is moving fast—and circling the drain.

Nathan Upchurch Nathan likes to cook vegan food, play the trombone, make incense, design stuff, muck about with free software and tech, and more. As a millennial, he dreams of owning a modest rural home in a sane country where he can make incense sticks in tranquility. To support Nathan’s writing (or simply help a filthy pinko leave the U.S.A.), you can donate on Liberapay.

Site Settings