We Were Promised Less Work and More Money. Here Is What AI Actually Delivered in 2026.

Last updated on May 29, 2026

The same lie was sold to farmers a century ago. It is being sold to knowledge workers right now. And the data does not care about your enthusiasm for the technology.

There is a pitch that never changes. The tools change. The decade changes. The PowerPoint slides change. But the promise stays exactly the same.

Less work. More money. Everybody wins.

They sold it with mechanised farming. People lost their jobs, worked longer hours, and had to retrain for roles that had not existed the year before. That was not the deal anyone signed up for.

Fast forward to 2026 and the script is identical. Except now the mechanism is generative AI, the boardroom pitch is “AI does the grunt work, humans do the thinking,” and the people getting squeezed are knowledge workers, executives, and business leaders who genuinely believed this time would be different.

It is not different. And I want to show you exactly why — with data, with enterprise case studies, with neuroscience, and with what it personally cost me when I went all in.

THE DATA IS NOT AMBIGUOUS

Here is what the research actually shows when you strip out the vendor case studies and the LinkedIn optimism:

Read that middle row again. Developers expected to go faster. They felt faster while using it. They went slower.

The work did not disappear. It shifted. Now instead of doing the job, you audit the machine. And the machine makes confident mistakes that you have to catch before they reach a client, a board, or a courtroom.

WHAT IT COST ME PERSONALLY

I went all in. Research, writing, strategy, client work — everything through AI. I have been doing this for five years, which is a long time considering what the landscape looked like before ChatGPT existed.

And what happened?

The writing started sounding like everything else. I was starting at the prompt instead of the problem.

There is a line in the research that names this precisely: cognitive surrender. The point at which you stop thinking and start reacting to what the model gives you. You default to its statistical biases, its reasoning structures, its rhythm. And slowly, without noticing it, you stop sounding like yourself.

The fix came from somewhere I did not expect. Brazilian jiu-jitsu. The Hayabusa. Thirty years of Tai Chi Neigong.

When you cannot prompt your way out of a choke — when someone is crushing you and you need to find a move or tap — your brain wakes back up. The problem is immediate, physical, and entirely your responsibility. No model is generating your escape route. No second opinion is available. You either think or you get submitted.

That contrast made the fog visible for the first time.

THE ENTERPRISE IS HITTING THE SAME WALL

This is not an individual problem. The biggest companies in the world are learning the same lesson at enormous cost.

Klarna fired 2,100 people in 2023, partnered with OpenAI, and ran its customer service almost entirely on conversational AI. By early 2025, customer complaints had spiked, satisfaction had fallen, and independent audits showed the chatbot was functioning as a rigid filter that pushed away high-value clients. The CEO publicly reversed course in May 2025, admitted the automation had compromised quality, and began rehiring human support layers.

Air Canada was held legally liable for financial damages after its customer-facing AI chatbot fabricated an unauthorised refund policy.

McDonald’s terminated its three-year automated drive-thru pilot with IBM after the AI repeatedly failed to interpret customer orders.

Microsoft deployed Claude Code to thousands of engineers in December 2025. By May 2026, it was quietly winding down those licences. The reason was financial. Agentic AI does not work on flat-rate SaaS pricing. It runs on token consumption, and a single engineering session can consume millions of tokens. Uber’s CTO revealed the company burned through its entire planned annual AI coding budget in four months, with individual engineers generating bills of $500 to $2,000 monthly.

A Harvard Business Review survey of over 1,000 executives found more than 600 admitted to cutting staff based on what they believed AI would be able to do — not what it could currently execute. Forrester reports 55% of those employers now regret it. Gartner projects 50% of AI-driven layoffs will be fully reversed.

And two days ago — on 26 May 2026 — Sam Altman walked back his own predictions on white-collar displacement in a speech to a banking audience in Sydney. His stated reason was direct: people want to deal with people on decisions that matter.

The architect of the disruption just told you there is a ceiling on it.

An MIT analysis confirmed the economics: AI is only financially superior to human labour on 25% of the white-collar tasks it was supposed to replace. The compute cost to scale the rest exceeds the salary of the human being replaced.

WHAT IT IS DOING TO YOUR BRAIN

This is where it gets uncomfortable.

The research category is called cognitive offloading, and the critical distinction it draws is between delegation and substitution.

Delegation is when you already know how to do something and you hand part of it to a tool. Your underlying cognitive pathways remain intact. Any decline is temporary and reversible.

Substitution is when you hand over a task you have never personally mastered. You skip the developmental step entirely. The result is cognitive foreclosure — a permanent state where you lack the internal baseline to evaluate, debug, or critique what the AI is producing.

A study on software developers learning a new coding library found that the AI-assisted group produced working code faster, but performed 17% worse on subsequent conceptual exams and was entirely unable to debug or explain what the AI had written. They had outsourced the learning itself.

The broader behavioural markers of what researchers are calling organisational cognitive atrophy look like this:

• Employees cannot initiate strategic thinking without first prompting a model
• Writing and analysis quality collapses when AI tools are unavailable
• Analysts cannot explain or defend the rationale behind AI-generated outputs
• Hallucinated or flawed outputs are accepted without critique because the interface is too frictionless to resist

The technical term for the end state is automation bias and epistemic passivity. The plain English version is: you stop thinking and call it efficiency.

THE FIX IS PHYSICAL. THE SCIENCE IS NOT SOFT.

Attention Restoration Theory identifies two types of human focus. Directed attention — active, analytical, fatigable. And involuntary or soft attention — passive, sensory, restorative.

Screen-heavy AI work forces the prefrontal cortex into sustained directed attention. When that system depletes, you get elevated cortisol, diminished working memory, and impaired emotional regulation. What researchers call “AI brain fry” is not metaphorical. It is a measurable physiological state.

The antidote is physical sovereignty — the deliberate return to embodied, modal-specific practice in real environments.

Aerobic physical activity and coordinated movement also trigger the release of Brain-Derived Neurotrophic Factor (BDNF) — the protein that promotes synaptic plasticity and neurogenesis. Outdoor physical activity with high cognitive demand generates measurable hippocampal growth. The brain structures depleted by digital overload are literally repaired by physical practice.

This is not wellness content. This is structural neurobiology.

FOUR DAILY PROTOCOLS THAT ACTUALLY WORK

These are not suggestions. They are the 20% of behavioural adjustments that produce 80% of the cognitive recovery.

1. The First Hour Sanctuary: Protect the first hour of your working day entirely from AI tools and digital dashboards. Write your three strategic priorities by hand. The mechanical act of handwriting engages distinct tactile motor pathways that promote deep conceptual synthesis and long-term retention. Start the day with independent cognitive planning, not a reaction to automated summaries.
2. The Offline Thinking Block: Establish a daily 60–90 minute block of unassisted, deep intellectual work. Offline. Paper and ink. Design architectures, outline proposals, read dense texts without the option to paste anything into a model. This prevents the prefrontal cortex from atrophying through continuous algorithmic prompting.
3. Architectural Friction: Consolidate all AI tools into a single nested folder on your desktop and phone. Remove shortcuts. Remove automatic browser tabs. This forces a conscious, deliberate choice to open the tool rather than a reflexive drift into it. Every interaction becomes intentional.
4. The Three-Prompt Boundary: Before you start any prompt sequence, write one explicit sentence stating the precise outcome required. If the model has not delivered it within three rounds, close the interface and execute the task yourself. This prevents the cognitive sink of endless prompt-refining and keeps the creative direction in your hands, not the model’s.

CURATION BEATS CONSUMPTION. EVERY TIME.

The highest-performing leaders in 2026 are running fewer tools, not more. The competitive advantage has shifted from tool adoption speed to curation discipline.

The problem was never the technology. The problem is continuous, undifferentiated consumption of AI outputs without a human thinking layer in between. Strip that layer out and you do not get efficiency — you get a faster version of someone else’s average.

Your judgment, built over decades of unassisted strategic execution, is the asset. Not your ability to prompt faster than the person in the next office.

Protect it accordingly.

If you want the deeper expert layer on this — the interviews, the analysis, the signal without the noise — check out Monday Influencer® at $19.95 a month. Steven J. Manning and I record a lot of interviews with people worth listening to.

And if this weekly briefing is useful, subscribe to Future Proof Leader — whether you want to read it, listen to it, or watch it, it is here every week.