Let's talk once more about the elephant in the executive suite. If we're honest (and honesty doesn't come easy under the industry mantra of the “war for talent” that's been drilled into us), there's a quiet, almost guilty sense of relief in the air whenever a senior manager can delegate a task to an LLM instead of to a junior employee.
Perfection isn't the point. The illusion that AI delivers finished, flawless results usually evaporates after the first week on the job. The problem is this: the fresh graduate doesn't deliver flawless results on the first attempt either. Both sides “hallucinate” in their own way.
So why is the junior currently losing this unofficial contest so mercilessly? The answer lies in an often ignored efficiency metric: the “Correction Effort”, the true cost of correction.
The anatomy of fallibility
Let's look at the nature of the mistakes that land on a senior's desk for correction.
The junior mistake usually stems from uncertainty, missing Tacit Knowledge (experience-based knowledge of a company's unwritten rules) or classic imposter syndrome: the reluctance to ask the right questions in time for fear of looking incompetent. The result is well-intentioned but incomplete output that often misses the core of the task (and I won't even get into the new problem: the junior using AI to deliver garbage).
The AI mistake is of an entirely different nature: an agent happily produces confident bullshit, weaves in plausible-sounding but fabricated nonsense, or stubbornly ignores three explicit exclusion criteria from the prompt.
For the senior, the consequence is identical in both cases: the actual work (reviewing, auditing, correcting) stays on his plate, because in the end he carries the responsibility. The decisive difference, however, lies in the iteration process that follows.
The tipping point paradox and the “Empathy Trap”
Why does the machine leave young talent standing in corporate reality? Essentially, three factors shift the socio-economic balance against the human:
- Iteration speed: If the first result is wrong, the AI generates “version 2” in about four seconds. The junior needs until tomorrow afternoon.
- Reliable availability: AI doesn't take unannounced days off, doesn't lose focus when the project gets tedious, and knows no loss of motivation.
Another weighty factor, though, is the removal of emotional friction. Just this spring, in my thoughts on the Empathy Trap, I shared a rather uncomfortable insight on this. In a heated debate about Business Process Optimization (BPO), this sentence slipped out of me:
“It's simply easier for me to work with AI agents, because I don't have to deal with human sensitivities.”
As painfully unempathetic as that sounds, it hits the cold core of the corporate optimisation imperative. Working with people constantly demands cognitive energy. You have to phrase feedback carefully, spare egos, motivate, stay politically correct. An agent, on the other hand, I can mentally just throw this at: “This is absolute garbage, start over from scratch, and leave out X.” No hurt feelings. No demotivation. No diplomatic tiptoeing.
In companies' pursuit of maximum efficiency, the very core of BPO, every human emotion, every frustration and every misunderstanding is pure friction loss. Technology removes that friction. The economic pull towards this “frictionless” state is as inescapable as the laws of thermodynamics. Maximum Efficiency requires the elimination of human emotion (another sentence that keeps echoing loudly).
The real-world “Friday afternoon test”
Nothing illustrates this paradox better than a classic “Friday afternoon test”.
It's Friday, 3:00 pm. A panicked client demands a short-notice, radical overhaul of a pitch deck or a software component by the evening.
How it used to go: The senior drags the young colleague out of the weekend he's already mentally started, manages his visible frustration on the call, wrestles with his own guilty conscience, and then at 6:00 pm still spends an hour rewriting 60% of the work (“because at this point it's just faster than explaining it again”).
How it goes today: The senior builds himself a brutal prompt workflow (if he knows how, but that's a topic for a future post), runs the AI's output through three merciless iteration loops in ten minutes (with zero regard for the AI's “feelings”), polishes the result and heads into the weekend at 4:00 pm. Efficiency triumphs over pedagogy.
The inevitable consequence: educational institutions as the new “senior forges”
From a corporate perspective, this leads us to a hard but unavoidable truth: No company can ignore the optimisation imperative in the long run. We can preach as many lofty corporate values as we like; at the end of the day, frictionless efficiency wins. Which simply means that classic, inefficient junior mentoring in business is dying a slow, quiet death.
Going forward, companies will be less and less willing and able to handle the laborious “seniorisation” of newcomers as part of daily business. It will simply no longer be economically justifiable to pay a senior to run pedagogical correction loops with a human when he can get the same done with a machine in a fraction of the time.
If business abolishes on-the-job training for the sake of optimisation, responsibility inevitably shifts to educational institutions.
Universities, business schools and training companies can no longer rest on handing over rough diamonds with a bare foundation of theoretical knowledge. The task of educational institutions must be to carry out the “seniorisation” described in the previous article while students are still on campus.
For educational institutions, this is exactly what becomes the ultimate Unique Selling Proposition (USP) of this decade: the universities that manage to rebuild their teaching so that their graduates no longer enter the market as “order takers” but as evaluating, critically thinking AI operators with systemic judgement will dominate the market.
Whoever delivers graduates to companies who are no longer an expensive “Correction Effort” risk solves one of the problems of modern management in the AI age. The era of “learning by doing” on the job is over. The era of “auditing by design” has to begin in the classroom. And that will be the topic of future posts, so stay tuned 😉
