The layoff spreadsheet just took a statistical slap in the face.
For months, the dominant narrative has been almost comfortably simple: artificial intelligence arrives, white-collar workers panic, junior roles disappear, companies capture productivity gains, and finance teams cut headcount with an Excel smile.
That narrative has one advantage: it is clear.
It also has one major weakness: it is far too tidy to describe reality.
A new analysis published by Ramp damages this linear story. By linking AI spending data with workforce data collected with Revelio Labs, Ramp researchers found that U.S. companies investing most intensely in AI grew their headcount by 10.2% in the two years following adoption. Even more interestingly, entry-level headcount grew by 12% among these high-intensity adopters (Ramp).
That seriously weakens the fable of the junior employee sacrificed on the altar of ChatGPT.
The Lazy Promise
The lazy promise was simple: install AI, reduce headcount, admire the margin.
On an executive committee slide, it looks beautiful.
In real life, it is often poor work architecture disguised as managerial courage.
AI does not automatically transform an organization. It amplifies what the organization already understands about how it works. A company able to analyze its workflows, irritants, decisions, dependencies and blind spots can use AI to increase execution capacity.
A company that does not understand how work actually gets done may confuse a job with a list of tasks.
That is where the damage begins.
A job is never merely a collection of micro-actions. It contains context, memory, judgment, experience, invisible trade-offs, weak signals, human relationships, perceived quality, responsibility and sometimes a simple ability to sense that something feels wrong.
When leadership eliminates a role before understanding that invisible depth, it is not automating.
It is amputating.
Advanced Companies Hire Differently
Ramp’s finding does not mean AI protects every job. It does not mean every sector will experience the same dynamic.
It says something more useful: companies adopting AI intensively are often already larger, more technical, better funded, faster and better organized. Ramp specifies that the observed gains are concentrated among high-intensity adopters, while lower-intensity adopters show no statistically significant change (Ramp Economics Lab).
In other words, AI does not reward incantation.
It rewards organizational capability.
My hypothesis is simple: the most advanced companies do not merely replace tasks. They redesign workflows, create new needs, increase execution speed and reinject that capacity into growth.
They do not use AI as a budget chainsaw.
They use it as process infrastructure.
That difference changes everything.
When a developer becomes faster, that developer does not necessarily disappear. They may produce more, test more, correct faster and ship more often.
When a marketing team automates part of production, it can also spend more time on strategy, differentiation, narrative and market signals.
When a finance team accelerates analysis, it can guide, detect, anticipate and advise more effectively.
Released capacity becomes performance only when the organization knows what to do with it.
The AI Layoff Boomerang
Other companies do the opposite.
They confuse automation with strategy.
They eliminate roles before understanding what those roles actually contained. Then they discover hidden costs: mediocre quality, content to rewrite, fragile decisions, degraded customer experience and managers overwhelmed by systems they do not master.
Robert Half reports that 54% of executives expect AI to lead to job growth rather than contraction over the next two years. The same source indicates that 32% of U.S. hiring managers who eliminated positions after implementing AI later had to add those roles, or very similar ones, back (Robert Half).
Gartner points in the same direction for customer service: by 2027, 50% of companies that attributed headcount reduction to AI will rehire staff to perform similar functions, sometimes under different job titles (Gartner).
The reason is clear: AI can handle part of the work, but part of the work does not make a complete job.
In customer service, for example, answering a simple question is only a fragment of the role. Handling frustration, sensing nuance, defusing tension, arbitrating an exception and preserving customer trust belong to another level of complexity.
Companies are discovering that cutting too fast can cost more than transforming properly.
Humans Are Cleaning Up Slop
Another interesting signal: humans are now being hired to clean up AI work.
The term “AI slop” refers to output generated at high speed, often acceptable on the surface but weak, repetitive, dull, approximate or poorly adapted to context. Media reports describe freelancers being hired to rework AI-generated content so it becomes usable, readable or simply acceptable (NBC News, NDTV).
Delicious.
The tool expected to eliminate work creates correction work.
The problem does not only come from the tool. It comes from naive use of the tool.
Asking AI to produce volume without direction, standards, editorial line, context or human validation creates quality debt. That debt must be paid. Sometimes quickly. Sometimes by someone more competent than the person who was replaced.
Slop is not technological destiny.
It is often the symptom of absent governance.
The Return of Philosophers
Another delicious narrative violation: philosophers are returning through the front door.
For years, some literary and philosophical backgrounds were treated with condescension by the “everything must be technical” crowd. Then generative AI made visible questions engineers alone cannot absorb: what is acceptable model behavior? What is an honest answer? What value should be preserved? How should organizations handle consciousness, intention, responsibility, manipulation and trust?
Business Insider reported that companies such as Anthropic and Google DeepMind are recruiting philosophy profiles to work on alignment, ethics and model behavior (Business Insider).
Daily Nous, which follows the philosophy profession, also covered the New York Times treatment of the topic, mentioning several philosophers working in or with AI-related organizations, including Robert Long, Geoff Keeling, Iason Gabriel, Patrick Butlin and Amanda Askell (Daily Nous).
The most advanced technology of the moment reminds us of an old truth: every important problem is not only technical.
Companies obsessed with code are discovering the weight of judgment.
Organizations fascinated by automation are rediscovering the value of critical thinking.
Leaders who wanted to replace humans with models are discovering they need humans to decide how those models should behave.
The Real Cost of Poorly Designed Automation
Orgvue published figures illustrating the difficulty of AI transformation. In 2025, 39% of surveyed leaders said they had made employees redundant as a result of deploying AI, and 55% of that group admitted they had made wrong redundancy decisions (Orgvue).
In 2026, Orgvue also reported that 32% of organizations that made layoffs based on AI cost-saving promises had to rehire staff, revealing a gap between technological ambition and operational understanding of work (Orgvue).
The most interesting part is not the number alone.
The most interesting part is the cause.
Companies do not get it wrong because AI is useless. They get it wrong because they deploy AI without understanding work.
They see tasks.
They miss the system.
They see costs.
They miss dependencies.
They see prompts.
They miss responsibility.
Poorly designed automation creates an illusion of productivity. It temporarily lowers the payroll line, then raises costs elsewhere: quality control, customer complaints, content rework, technical debt, loss of tacit knowledge, managerial overload, demotivation and loss of trust.
AI as Process Innovation
That is precisely why I treat AI as process innovation in my Innovational Intelligence system: clear objective, multidisciplinary team, experimentation, learning, governance, sentiment measurement and progressive deployment (my book, chapter 14).
AI is not merely a tool.
It is a transformation of how work is designed, executed, controlled and valued.
In my approach, the first step is not to ask: “How many jobs can we cut?”
The first step is to ask: “What new capability do we want to create?”
Then come the useful questions:
Which function should be augmented?
Which workflows should be redesigned?
Which irritants should disappear?
Which roles should evolve?
Which risks should be controlled?
Which talents should be trained?
Which indicators will measure quality, speed, trust, customer experience and learning?
Which guardrails will protect the organization from the illusion of performance?
A company asking these questions gives itself a chance to use AI as a growth lever.
A company that jumps directly to headcount reduction may prepare itself to rehire later, with less trust, less internal competence and more cultural damage.
The Metric That Changes Everything
In AI adoption, the central metric is no longer: “How many jobs did we cut?”
The useful metric becomes: what new capability can the organization create?
Capacity to serve faster.
Capacity to decide better.
Capacity to learn faster.
Capacity to personalize without degrading quality.
Capacity to produce more without producing slop.
Capacity to free humans from low-value tasks and reposition them on decisive ones.
Capacity to create new roles.
Capacity to bring together engineers, business teams, lawyers, designers, HR, salespeople, philosophers, managers and executives.
Used well, AI does not make humans decorative.
It makes work design strategic.
The Trap of the Rushed Executive
The rushed executive sees AI as a shortcut.
The lucid executive sees it as a revealer.
AI reveals absurd processes, silos, badly defined roles, poorly governed data, undocumented decisions, blurry responsibilities, undertrained managers and cultures unable to learn.
AI placed on organizational chaos does not create an intelligent organization.
It creates faster chaos.
That is why I talk less about “digital transformation” and more about operational lucidity. The challenge is not to place a chatbot on an open wound. The challenge is to understand work, augment what deserves to be augmented, experiment, measure, correct and then deploy.
Replacing without understanding destroys competence.
Augmenting with method creates performance.
What Leaders Should Do Now
I see three priorities.
First, map real work. Not job descriptions. Not org charts. Real work: decisions, exceptions, trade-offs, dependencies, irritants, duplicates, risks, expected quality and tacit knowledge.
Second, select AI use cases through a process innovation lens. A good use case is not only one that saves time. It improves speed, quality, experience, learning and robustness at the same time.
Third, protect the human who becomes more valuable. AI makes some tasks less scarce. It makes judgment, responsibility, critical thinking, creativity, communication and learning capacity far more valuable.
That is why juniors are not necessarily doomed.
They can become the first operational natives of this new way of working, provided companies do not treat them as adjustment variables.
That is why philosophers are becoming interesting.
They bring vocabulary and disciplined thinking to topics technology has made urgent.
That is why slop editors are becoming necessary.
They remind us that producing fast is never enough.
The Right Provocation
The next time a slide promises headcount reduction through AI, ask one simple question:
What new capability are we going to create?
When nobody can answer, you are not looking at an AI strategy.
You are looking at impoverishment disguised as modernization.
👉 Which function in your company should be augmented by AI before being stupidly automated?
Yes, I also bring this topic to my keynotes, workshops and advisory work, with fewer “digital transformation” slides and more operational lucidity.
References
- (Ramp) = https://ramp.com/data/ai-jobs-impact
- (Ramp Economics Lab) = https://ramp.com/data/heavy-ai-adopters-hire-more
- (Robert Half) = https://www.roberthalf.com/us/en/insights/landing-job/job-seeking-these-sectors-are-hiring-now
- (Gartner) = https://www.gartner.com/en/newsroom/press-releases/2026-02-03-gartner-predicts-half-of-companies-that-cut-customer-service-staff-due-to-ai-will-rehire-by-2027
- (Orgvue) = https://www.orgvue.com/news/55-of-businesses-admit-wrong-decisions-in-making-employees-redundant-when-bringing-ai-into-the-workforce/
- (Orgvue) = https://www.orgvue.com/news/new-research-exposes-the-complexity-of-deploying-ai-systemsin-the-workforce/
- (Business Insider) = https://www.businessinsider.com/ai-job-market-careers-philosophy-majors-google-anthropic-2026-4
- (Daily Nous) = https://dailynous.com/2026/07/05/nyt-the-revenge-of-the-philosophy-major/
- (NBC News) = https://www.nbcnews.com/tech/tech-news/humans-hired-to-fix-ai-slop-rcna225969
- (NDTV) = https://www.ndtv.com/offbeat/humans-in-demand-to-clean-up-ais-mess-amid-job-loss-fears
