I oversee AI strategy for a Fortune 200.
My title didn't exist three years ago.
Neither did 70% of the titles like mine.
Chief AI Officer appointments were up 70% year over year in 2024.
Most of us leave within eighteen months.
The position has been described as "Chief of Nothing."
Not by critics.
By the people who hold it.
One of my peers described the job as "trying to build a Formula 1 team while only being able to offer bicycle mechanic salaries."
He was fired four months later.
He was replaced within the week.
The replacement was also fired.
I've outlasted both of them.
Not because I've delivered more.
Because I present better.
In 2023, McKinsey told the board that AI would boost productivity by 40%.
The board approved $140 million in AI spending.
In 2024, the industry invested $250 billion.
In 2026, 6,000 CEOs were surveyed across four countries.
90% said AI had no impact on employment.
90% said AI had no impact on productivity.
The average executive uses AI 1.5 hours per week.
25% don't use it at all.
$250 billion.
Ninety minutes a week.
MIT studied 52 organizations that had spent a combined $30 to $40 billion on generative AI.
95% achieved zero return on investment.
Zero.
Not low returns.
Not disappointing returns.
The number between negative one and one.
I presented these findings to the board.
I titled the slide "Opportunity for Differentiation."
Sequoia Capital asked the question in June 2024.
David Cahn called it "AI's $600 Billion Question."
AI companies need to generate $600 billion in annual revenue to justify current infrastructure spending.
They are generating a fraction of that.
Cahn's exact words: "We need to make sure not to believe in the delusion that has now spread from Silicon Valley to the rest of the country, and indeed the world."
I printed that quote.
I hung it above my desk.
As a reminder to believe harder.
Our company bought Microsoft 365 Copilot for 12,000 employees.
$30 per user per month.
$4.32 million a year.
Microsoft's total Copilot adoption rate is 3.33%.
15 million paid seats out of 450 million commercial subscribers.
96.67% of the people who could use it don't pay for it.
Satya Nadella's response to this adoption rate was that people are using Copilot "a lot."
TechCrunch put "insists" in the headline.
Nadella made $96.5 million last year.
A 22% raise.
The raise was not tied to Copilot adoption.
The raise was tied to the stock price.
The stock price was tied to the AI narrative.
The AI narrative was tied to the pitch deck.
The pitch deck is the product.
I know this because I've watched it happen.
Goldman Sachs published a report in June 2024 titled "Gen AI: Too Much Spend, Too Little Benefit?"
The question mark is doing a lot of work in that title.
Their own CEO, David Solomon, said in October: "I expect there to be a lot of capital that was deployed that doesn't deliver returns."
Jeff Bezos, same month: "Kind of an industrial bubble."
Sam Altman, same month: "People will overinvest and lose money."
These three men run companies that are investing $602 billion in AI infrastructure in 2026.
They said it's a bubble.
Then they inflated it.
I mention this at no meetings.
Gallup surveyed 22,000 Americans.
49% of U.S. workers have never used AI.
Not "rarely use."
Never.
The majority of the American workforce has not touched the technology that $602 billion is being spent on.
Slack surveyed 10,000 desk workers.
Two-thirds have never used AI at work.
93% don't trust AI outputs.
When workers were asked what they'd do with time saved by AI, the number one answer was "more admin."
Not innovate.
Not create.
More admin.
We are spending $602 billion so people can do more admin.
The Workday study found that 37% of time saved by AI is immediately consumed by rework.
Employees spend the saved time correcting, rewriting, and contextualizing what the AI produced.
The net savings: approximately 1.5 weeks per year.
Minus the time spent in the mandatory AI training sessions.
Minus the time spent in the meetings about the AI training sessions.
Minus the time spent reporting on the meetings about the AI training sessions.
I have not calculated the net.
I suspect it's negative.
I am not going to calculate it.
Our last all-hands featured a live AI demo.
It went well because I rehearsed it eleven times and pre-loaded the prompts.
Google was less careful.
Their Gemini launch video hit one million views.
It showed the AI responding in real time to voice and visual cues.
The entire demo was faked.
Still images. Text prompts. Edited together.
Google's defense: the video "shows real outputs" and they "made a few edits."
A VP of Research later clarified it showed what the experience "could look like."
"Could" is doing the same work as Goldman's question mark.
GPT-4 claimed to score in the 90th percentile on the bar exam.
Independent researchers replicated the study.
Actual score: 15th percentile.
The gap between the claim and the reality is 75 percentile points.
I have seen gaps like this before.
They're called pitch decks.
Amazon launched Amazon Go.
"Just Walk Out" technology.
AI-powered shopping. No cashiers. No checkout.
The AI was thousands of workers in India watching cameras.
Because the models didn't work.
Amazon called it AI.
The workers in India called it a job.
The workers in India were cheaper than the AI.
This fact appears in no pitch deck.
Klarna fired the equivalent of 700 customer service agents.
Replaced them with an AI chatbot.
The CEO bragged about it at conferences.
Revenue per employee hit nearly $1 million.
Then the chatbot started delivering lower quality.
The CEO's exact words: "Cost unfortunately seems to have been a too predominant evaluation factor."
Klarna started rehiring humans.
They reassigned software engineers and marketers to answer phones.
When asked if they had reversed their AI strategy, they said they had "not reversed or scaled back" their AI strategy.
They had reversed their AI strategy.
Salesforce cut 4,000 customer support jobs in September 2025.
Marc Benioff said: "I need less heads."
In January, Salesforce had already cut 1,000 jobs while hiring salespeople to sell the AI that replaced the 1,000.
By February 2026, they cut the Agentforce AI product team itself.
The team that built the AI that replaced the humans was replaced.
They hired six new executives to oversee the replacement of the replacement.
At the AI for Good Global Summit, Benioff had promised "radical augmentation" rather than job elimination.
He then eliminated 4,000 jobs.
The word "augmentation" and the number "4,000" were in the same fiscal year.
There is now an SEC enforcement unit specifically for AI-washing.
A company called Nate claimed its shopping app used "proprietary AI" with a 93% to 97% automation rate.
The actual automation rate was zero percent.
Zero.
The app was operated by hundreds of contract workers in a call center in the Philippines.
They raised $42 million.
The CEO directed employees to prioritize processing test transactions from potential investors.
He is facing two counts of fraud, each carrying a maximum of twenty years.
Another firm, Joonko, raised $21 million on false AI claims.
The SEC called it "old-school fraud using new-school buzzwords."
A cannabis company called Flora Growth rebranded as "ZeroStack Corp" and pivoted to "AI-focused asset management."
Their market cap: $6.61 million.
Their EBITDA: negative $13.71 million.
The CEO called it "a natural progression."
From losing money on weed to losing money on AI.
Naturally.
My board has asked me for the ROI on our AI investment.
I told them we're in the "investment phase."
Erik Brynjolfsson at Stanford calls the next part the "harvest phase."
I like that framing.
The harvest phase is always next quarter.
42% of companies abandoned most of their AI initiatives in 2025.
Up from 17% in 2024.
46% of AI proof-of-concepts die before reaching production.
One in fifty AI investments delivers transformational value.
Gartner said that.
I cite Gartner in every board presentation.
I do not cite the one-in-fifty number.
I cite the one that says the market will be worth $1.3 trillion by 2032.
Both numbers are from Gartner.
I choose which one to present.
That is my job.
76% of C-suite executives believe AI saves them more than 4 hours per week.
40% of workers say AI saves them no time at all.
Only 32% of non-managers have access to AI tools.
Only 27% received training.
The executives who believe AI saves 4 hours per week are the ones with access, training, and assistants who do the prompting for them.
They are measuring their assistant's productivity.
And calling it artificial intelligence.
Mustafa Suleyman, Microsoft's AI chief, said in February 2026 that AI will replace "most, if not all, professional tasks" within 12 to 18 months.
Dario Amodei, Anthropic's CEO, said AI could eliminate half of all entry-level white-collar jobs and spike unemployment to 10-20%.
He also estimated a 25% chance it goes "really, really badly."
These are the people building it.
One of them says it will replace everyone.
The other says there's a one-in-four chance of catastrophe.
Neither of them has stopped.
I was asked to present at our internal AI summit last quarter.
The talk was called "Accelerating Value Through Enterprise AI."
It was thirty minutes.
I used fourteen slides.
Eleven of them contained the word "transformative."
None of them contained the word "zero."
I got a standing ovation.
Not for the results.
For the confidence.
The confidence is the product.
My performance review says I have "successfully positioned the organization for AI-driven growth."
I have not delivered AI-driven growth.
I have delivered AI-driven positioning.
These are different things.
They pay the same.
I'll make SVP by year end.
The pitch deck is on slide 1 of every board meeting.
The results are on slide 47.
Nobody gets to slide 47.
$602 billion.
Ninety minutes a week.
The 40% improvement was the pitch deck.
The pitch deck is the product.
And business is booming.
English
