David Arnoux recently shared something that caught my attention: "I think we need to talk about this more. I don't think we're ready for what's coming." He went further, describing a reality he sees daily with founders and boards: "You can now do the work of a hundred 'white collar' people with one person."

That combination of urgency and specificity is what makes his post stick. David is not arguing for a world without humans. He is arguing for a world with far fewer humans required for many knowledge workflows. And that shift is already showing up in the most unforgiving place possible: the metrics boards care about.

The board-level pressure behind the AI wave

David Arnoux pointed out that boards are asking for "better revenue per employee" and "opex improvement." That is not abstract. It translates into very concrete decisions:

Fewer net new hires for the same growth targets
Consolidation of roles (one person covers what used to be three)
A bias toward systems and automation over headcount

If you are inside a company today, the immediate risk is not always layoffs. David’s nuance matters here: the bigger erosion starts with "new jobs, new positions" that simply do not get created.

In other words, the job market can weaken without dramatic headlines. It can weaken quietly, through missing rungs on the ladder.

"One person can do the work of 10, 20, 50 people"

A key part of David’s argument is that AI is no longer limited to isolated tasks like drafting an email. The unlock is when tools can connect into your real environment: your repository, CRM, CMS, internal knowledge, and the way your business actually works.

He referenced Claude Code as a catalyst, describing a setup where you "plug it into your repository" and "connect it to your CRM" and then "build up the agents and workflows." The implication is bigger than coding help. It is operational leverage.

"One person can do the work of 10, 20, 50 people. And the work is good."

The phrase "and the work is good" is doing a lot of work. Many skeptics assume AI output is low quality, so humans must review everything, so productivity gains stay modest. David is describing the opposite: once you have a solid foundation, quality can be consistent enough that the limiting factor becomes coordination, not capability.

The real bottleneck: data unification and architecture

One of the most important lines in David’s post is also the least glamorous: "Apart from a few people strong on database unification and architecture, I'm not seeing" the new jobs.

This frames a sobering point about the near future of knowledge work:

The high leverage work is building the layer that makes agents reliable
The repetitive work that used to train juniors gets automated first
The remaining human roles skew senior and systems-oriented

If you have ever tried to automate a revenue operation, you already know why. The tool is rarely the hard part. The hard part is getting clean definitions (what is a lead, what is a qualified opportunity), consistent fields, versioned content, permissions, and governance.

AI agents thrive on coherent systems. When the system is messy, they hallucinate, mis-route, or do the wrong thing confidently. So companies that invest in the unification layer get compounding returns, while everyone else gets a pile of disconnected copilots.

Entry-level roles are the first shock absorber

David Arnoux wrote that people criticized him months ago when he said "entry level hiring was getting hit" and that now "it's happening." This matches what many teams are quietly doing:

Hiring one strong generalist instead of two juniors
Expecting new hires to be tool-fluent on day one
Using AI to do the "first draft" work that used to be delegated

Why does this matter so much? Because entry-level roles are not only labor. They are training infrastructure. They are how organizations create future senior talent. If those roles thin out, you do not just get short-term savings. You change the shape of the future workforce.

A hard question: what are the new jobs this time?

David asked the uncomfortable question directly: every tech revolution created new jobs, "but what are the new jobs this time?"

There will be new roles, but they may be:

Smaller in headcount than the roles they replace
Concentrated in companies that successfully rebuild around automation
More technical and systems-heavy than many people expect

That is why the transition feels harsh. It is not that work disappears. It is that the distribution of work changes faster than people can retrain.

"Computers interface with computers": the two-internets idea

One of David’s most original points is philosophical: "We were always just factory workers for computer interfaces." He argues the UI existed mainly so humans could translate intent into machine action. Now, "computers interface with computers," and that changes everything.

He predicts "two internets":

A human-to-human internet for social media, entertainment, and doom scrolling
An agent-to-agent internet for vendor selection and deep technical work

This is not science fiction. You can already see hints of it:

AI agents researching products, comparing options, drafting shortlists
Automated outreach, automated qualification, automated follow-up
Machine-readable content (APIs, schemas, feeds) mattering more than glossy webpages

If the buyer’s first interaction is an agent, then the game shifts from persuasion to parseability and proof. Clear pricing, strong documentation, verified security posture, and structured data become growth levers.

Atoms vs bits: where humans may be safer (for now)

David’s practical advice was blunt: "Get on the bandwagon now and be cutting edge..or..get into longer term deep projects." He also pointed to "building (actual) bridges" and noted we need more welders, mechanical engineering, and physical-world work.

The "atoms versus bits" framing is useful. Many digital workflows compress easily because they are information transformations. Physical workflows resist compression because they involve:

Safety and liability
Complex environments and edge cases
Hardware costs, maintenance, and regulation

He also added a values-based point that deserves repeating: teachers, doctors, and care workers should be paid far more. Even if AI supports these professions, the human relationship and accountability remain central.

What to do if you are a founder or GTM leader

If David is right, leaders should treat AI as a structural change, not a tool rollout.

1) Measure leverage, not activity

Revenue per employee is the scoreboard boards are watching. Build dashboards that reflect:

Output per role (pipeline created, content shipped, tickets resolved)
Cycle time (from idea to production)
Quality (win rates, churn, incident rates)

2) Build the unification layer first

Before you "add agents," fix the substrate:

Standardize fields and definitions across systems
Centralize documentation and decision logs
Implement permissions and audit trails

3) Redesign roles around workflows

Instead of hiring for tasks, hire for ownership of end-to-end workflows. The new bar is often:

Can this person orchestrate tools and agents?
Can they validate outputs and spot failure modes?
Can they improve the system over time?

What to do if you are early-career

The most actionable interpretation of David’s warning is: do not compete with the default AI workflow.

Become the person who can set up the workflow (data, prompts, guardrails, evaluation)
Pick deep projects where domain knowledge compounds (security, healthcare ops, compliance)
Build proof of work publicly (small tools, automations, repos, case studies)

And if your instincts pull you toward the physical world, take that seriously. The prestige economy may lag reality for a while, but demand is demand.

A reset, whether we like it or not

David ended with a balanced uncertainty: maybe it is "natural evolution," maybe he is in a bubble, but "it's the transition that hurts." I agree with that framing.

The argument is not that humans become irrelevant. It is that the old shape of white-collar work, the endless sifting through Excel, building PowerPoints, and staring at charts, may have been a temporary interface era.

If computers can talk to computers, our job becomes choosing the goals, designing the systems, and taking responsibility for outcomes. That is fewer people doing higher-leverage work, plus many people doing deeply human or deeply physical work.

This blog post expands on a viral LinkedIn post by David Arnoux, Helping GTM Leaders & Founders Grow With GTM x AI | Fractional CxO | Building Linkedin Tools @ humanoidz.ai. View the original LinkedIn post →