👋 Hi, it’s Maurizio. Welcome to my newsletter on everything professional services.
To find the capability currently blocking your system, and get an hyper-personalized plan for your performance transformation → Get Consulting Edge
Three things can be true at the same time:
AI will replace jobs,
AI is a source of slop generation,
Thanks to AI, there will be more jobs than today.
This is only apparently a paradox.
The moment you understand labor not as a static zero-sum system, but as a dynamic economy driven by moving marginal costs and expanding human ambition, is the moment you realize the combination of the three statement is the most likely outcome.
AI will replace jobs
The argument that artificial intelligence will replace jobs often meets a common defense:
No! A job is not merely an arithmetic sum of discrete tasks.
Viewed from the inside, human employment is rich with unquantifiable dimensions that resist any sort of algorithmic capture. A professional role encompasses tacit knowledge, political instinct, interpersonal empathy, institutional memory, and, importantly, legal and moral accountability.
An administrative assistant, for example, manages office dynamics, protects executive attention, gauges urgency through subtle behavioral cues, and serves as an informal communications hub. Reducing any job to a checklist of mechanical functions ignores the social and contextual work that keeps organizations operating smoothly.
Yet acknowledging this complexity cannot obscure an inescapable economic reality: in every major technological revolution, certain jobs have indeed vanished.
No occupation is granted permanent immunity.
Historically, this displacement occurs because it commoditizes the primary function that justified their salary.
The nineteenth-century typist pool, the early twentieth-century telephone switchboard operator, the mid-century draftsman, and the bank ledger keeper were not mindless automata. Nevertheless, their primary economic utility rested upon mediation: transcribing, patching circuits, drawing lines to scale, and balancing books.
Once mechanical switches, carbon paper, xerography, computer-aided design, and digital databases brought the cost of that core mediation close to zero, organizations reorganized around the new tool.
The intangible elements of those vanished roles did not save them. Instead, those responsibilities were either absorbed by remaining staff, simplified by better interface design, or deemed superfluous relative to the dramatic efficiency gains of automation.
Artificial intelligence initiates an identical structural pressure, this time targeting routine intellectual execution.
As I have said before, the danger to white-collar employment lies in the mechanism of unbundling, with LLMs and autonomous agents now able to instantaneously draft legal briefs, translate code across programming languages, summarize financial statements, or handle first-tier customer grievances.
Even if an AI cannot assume legal liability, build deep client trust, or comprehend high-level strategy, organizations do not need an entire department of junior analysts simply to preserve those qualitative virtues!
Instead, firms will unbundle the role.
They will assign the high-volume, green-path execution to software, while consolidating the residual human responsibilities into a much smaller tier of senior supervisors.
To assert that AI will replace jobs is simply a sober recognition of economic gravity.
A job may be richer and more holistic than the tasks listed on its description, but when the foundational tasks that justify its headcount are automated away, the social and institutional intangibles cannot carry the payroll alone.
As with every historical transition, some titles will survive by evolving, but many others will follow the switchboard operator into economic obsolescence.
AI is a source of slop generation
The second reality is that LLMs are engines of industrial-scale mediocrity.
Going back to basics, generative models do not comprehend reality, test hypotheses against physical constraints, or possess internal models of objective truth. They are statistical prediction engines, engineered to generate sequences of tokens that maximize surface plausibility based on patterns extracted from their training corpora (which is all stuff produced by humans).
Historically, bad information easily showed its unreliability through visible flaws: clumsy syntax, broken grammar, disjointed logic, or bad formatting. Human beings developed heuristics that used stylistic polish as a proxy for rigor.
Generative models have cleanly severed this connection (hell, I am not a native English speaker so all my writing gets a LLM pass to check grammar and mistakes - thank the LLM for that!)
An LLM can construct an essay, an architectural review, or a legal brief that possesses immaculate cadence, an authoritative voice, and flawless grammar, while simultaneously asserting total factual bullshit, citing non-existent court precedents, and fabricating statistical data out of thin air.
To a novice or an outsider, this is nearly indistinguishable from genuine mastery. But the moment a true domain expert inspects the output, that thinking goes away quickly...
A senior software engineer recognizes that a generated function, while visually elegant, introduces subtle race conditions and architectural anti-patterns that will create issues under real-world server loads.
A seasoned physician notices that the synthesized diagnostic summary conflates correlation with causality and overlooks rare contraindications that would alter clinical care.
A research historian instantly catches the anachronism or realizes that a cited archival source simply does not exist.
At the end of the day, expertise is the accumulated, intuitive grasp of boundary conditions, structural exceptions, edge cases, knowing not just what is usually said, but what can never be true.
Generative AI possesses vast familiarity with the median, but it has no authentic understanding of reality’s hard edges. Because it operates on probability, it is congenitally incapable of knowing when it does not know.
The existential problem for modern organizations and public discourse is that while producing this synthetic output costs fractions of a cent and takes mere seconds, evaluating it is exorbitantly expensive.
The information ecosystem is now confronted with an algorithmic variation of Brandolini’s Law:
The amount of energy needed to refute or untangle automated nonsense is an order of magnitude larger than the energy required to generate it.
Look at what’s going on:
Academic journals are overwhelmed by thousands of synthetically generated papers accompanied by fabricated peer reviews.
Open-source software repositories are inundated with pull requests generated by automated agents that introduce phantom dependencies and invisible security vulnerabilities.
Corporate inboxes, hiring pipelines, and customer support channels are suffocated by low-effort prose designed solely to game algorithmic filters.
This deluge threatens to contaminate the very foundations of future intelligence itself.
As the open web becomes saturated with auto-generated text and synthetic imagery, subsequent generations of models risk training on their own regurgitated output: an algorithmic feedback loop known as model collapse.
I believe this hyper-production of slop radically inverts the value proposition of knowledge work.
Value migrates upstream to the domain experts who possess the contextual depth and critical skepticism required to audit, debug, filter, and reject synthetic garbage.
The modern economy is therefore realistically headed toward an information landscape choked with plausible falsehoods and synthetic clutter.
Surviving it will depend on who retains the human expertise required to tell the difference between insight and nonsense.
Thanks to AI, there will be more jobs than today
The horizon of human labor is an expanding frontier.
The conviction that machines will permanently exhaust the need for human work rests on one of economics’ most persistent intellectual fallacies: the belief that the total amount of useful work in the world is fixed.
AI unleashes an explosion of latent demand while creating unprecedented organizational complexity that forces the economy to invent entirely new categories of human endeavor.
The baseline mechanics of this phenomenon begin with the paradox observed by the Victorian economist William Stanley Jevons in 1865.
When James Watt made the steam engine vastly more fuel-efficient, observers assumed the world would burn less coal. The opposite occurred! Cheaper mechanical energy made steam power economically viable across thousands of applications that had previously been unthinkable, causing aggregate coal consumption to skyrocket.
What is happening today?
Well, the cost of writing code, analyzing datasets, designing molecules, or simulating physical stress tests is dropping by orders of magnitude, but, as this happens, society does not bank the savings and keep demand flat.
We do not simply build the same software for less money; we make software ubiquitous across every physical artifact, biological inquiry, and civic process on Earth.
The demand for computation, integration, and problem-solving is functionally infinite! And if we keep lowering the energy required to attempt ambitious projects, then it is reasonable to expect that AI will cause the total volume of attempted ambition to multiply exponentially.
Yet Jevons Paradox is merely the entry point to a far more profound transformation.
The primary reason public discourse continually succumbs to fatalistic predictions of technological unemployment is a structural failure of imagination.
We suffer from an intellectual asymmetry: the jobs that will be destroyed are concrete, visible, and historic, whereas the jobs that will be created are abstract, invisible, and reliant on an ecosystem that does not yet exist.
In 1800, when 80%+ of the labor force was bound to agricultural toil, it was impossible for an observer to conceive of cybersecurity engineers, pediatric physical therapists, satellite telecommunications directors, or avionics specialists.
If an observer could only imagine human utility through the prism of farming, the tractor appeared to be an instrument of terminal starvation.
When my grandfather (who died at 99 a few years ago) was 18, could have never imagined that, during his lifetime, his grandchild (me) would have talked to his daughter (my mother) through a box, while each of us sits in two different continents.
To deduce that human labor will end simply because our present vocabulary cannot name the job descriptions of 2060 is actually a form of hubris, as we are assuming our contemporary imagination marks the outer boundary of human civilization.
Think about the modern software stack.
60 years ago, computers were programmed using punch cards. Compilers, high-level languages, and pre-built frameworks successively automated what was once the tedious, artisanal work of manual assembly translation. By the logic of task displacement, the world should have required fewer programmers with each leap in automation. Instead, lowering the effort of code creation allowed software systems to become millions of times larger, more interconnected, and more delicate.
That immense scale introduced radically new vulnerabilities, operational dependencies, and architectural dilemmas. It created the very fields (DevOps, cloud security, telemetry engineering, data governance) that currently employ millions.
By the way, this is why “consulting” is such a resilient field: when a solution to a problem appears, three new problems that didn’t exist before become evident that need solutions for.
As generative AI automates baseline code and prose, the systems humans build will balloon in scale and ambition.
We will attempt planetary-scale energy grids, personalized medicine down to the individual genome, real-time climate modeling, hyper-complex logistics networks, and far more innovative stuff I can’t even name right now.
Simultaneously, value obeys economic gravity: it flees what becomes abundant and concentrates entirely in what remains scarce.
If automated generation of synthetic text, imagery, and code will become so ubiquitous and essentially free as I imagine, pure computational output will cease to be a luxury good.
What becomes scarce, and therefore immensely valuable, are the once incomputable assets of human existence: authenticity, presence, context, legal and moral accountability, institutional trust, discernment (by the way, I created Consulting Edge as a way for professionals to measure, benchmark and improve on these essential skills before it is too late).
Economic activity will re-organize around these scarcities.
The upcoming economy will be likely dominated by bespoke craft, interpersonal care, rigorous investigative and forensic oversight, ethical governance, community architecture, mostly direct human-to-human coordination.
Roles that were once considered peripheral or economically unviable will move to the financial center stage, subsidized by the astronomical productivity gains of underlying automated infrastructure.
Stay optimistic
This hyper-democratization of agency will fragment the landscape into an archipelago of micro-enterprises, niche research collectives, specialized service boutiques and, of course, independent creators.
Millions of people who were previously passive employees within rigid corporate bureaucracies will become operators, creators, and leaders of entirely new initiatives, each generating its own micro-economy of human exchange and specialized labor.
The three truths of artificial intelligence are not contradictory.
AI will unbundle and destroy established jobs by commoditizing routine execution. It will inundate our shared informational ecosystem with a mountain of synthetic slop that demands human discernment to filter and correct. And precisely because it reduces the cost of execution while multiplying complexity, it will expand the total surface area of human civilization, unlocking new domains of inquiry and enterprise.
Work is not a finite pool of drudgery to be drained dry by algorithms!
Work is the formal expression of humanity’s boundless desire to understand, repair, elevate, and explore reality.
Machines may conquer the median of execution, but in doing so, they will clear the ground for human beings to imagine new and more audacious work.
As usual, if you enjoy reading Consulting Intel, please do me a favor: spread the word and share this post.
👋
👀 Links of interest
A few corners of the internet you may find interesting:
This thread on my X’s profile resonated with a lot of people. What feels like effortless intuition to one person is almost always a sequence of essentially subconscious steps to everyone else:
Maurizio@themgmtconsultWhen people reach the upper tiers of an organization, there is an unspoken assumption that circulates in the corridors about them:11:00 AM · Sep 1, 2026 · 32.8K Views4 Replies · 3 Reposts · 45 Likes
Have you looked into the Leaders Toolkit? It is a deck of 52 tools, frameworks and mental models to make you a better leader (use code CONSULTANT10 for 10% off);
The Consulting Intel private Discord group with 250+ global members is where consultants meet to discuss and support each other (it’s free).
Very timely, there's an article in this week's The Economist specifically about how AI is creating a mini-boom, not just in data centre infrastructure but even in white collar jobs that were supposed to be going away (e.g. paralegals).
Totally agree. In my consulting work and while I’m using AI extensively to create ‘first drafts’ I’m finding that while editing the drafts I’m having myself force fit certain ideas into what the AI has recommended. In other cases where I use AI to help me group common themes, a lot of meaning is lost. Believe a big challenge now is AI over-dependency that is limiting our creative and in depth thinking.