Everyone is blaming AI for the layoffs, and the smartest people building the stuff are not buying it. Jensen Huang called the excuse lazy. Demis Hassabis of DeepMind called it dumb. Take-Two’s Strauss Zelnick said the quiet part out loud: the big companies who fired thousands and pinned it on AI “were not telling the truth.”

So why are they really doing it? Follow the money. Meta is spending up to $145 billion on AI this year against a payroll a fraction of that size, which means the layoffs are not paying for the AI at all, they are making a wildly expensive bet look responsible to analysts who would otherwise ask hard questions. Calling overcapacity “AI” is gaslighting the exact people you need to make the tool work.

The future does not belong to the boards bragging about doing more with less. It belongs to the ones brave enough to do more with more. IBM is tripling its US entry-level hiring. Benioff hired 1,000 graduates to build the AI itself.

I wrote down the whole ugly mechanism, the circular money, the entry-level wipeout, the executives hiding behind a single word.

Read it here: https://heliade.net/ai-layoffs-excuse-more-with-more/

What was the real reason behind the last round of cuts you watched happen?

Which makes it deeply inconvenient that Pope Leo XIV has just written one of the sharper AI texts I have read this year.

Magnifica Humanitas should have been Vatican fog around a policy memo. Instead, the thing has teeth. Leo places AI beside Rerum Novarum, the 1891 encyclical where the Church looked at industrial capitalism and remembered, eventually, that workers were human beings rather than furnace accessories. Same move now: new machine; old question. Who pays with their body when progress needs fuel?

The scene around the document gives the game away. Christopher Olah, co-founder of Anthropic, stood in the Vatican and said frontier AI labs operate inside incentives that can conflict with doing the right thing. Commercial survival. Race pressure. Geopolitics. Pride. Ambition. He asked for critics outside those incentives, moral voices the labs cannot bend.

A man near the controls admitted the control room is insufficient. Wow. Silicon Valley spends its days pretending it escaped religion while speaking almost nothing else. Salvation has a model name. Apocalypse has a probability chart. Confession arrives as a model card. Prayer has a prompt window. Saints have podcast microphones and unusually liquid compensation.

Then Rome enters the conversation and gives the materialist lecture. (pun intended). The Pope talks about data centres, water, energy, miners, content moderators, children, labour dignity, platform concentration, public services, schools, families, kill chains, and the people buried under the word “cloud.” The AI labs talk about mystery, model welfare, inner states, superintelligence, geopolitical destiny, and whether machines may one day deserve moral consideration.

That inversion is bizarre. The spiritual leader is pointing at matter. The materialists are selling metaphysics. Somewhere, an irony meter catches fire and asks for a sabbatical.

Some in the AI safety world complained that Leo did not go hard enough on AGI and superintelligence. The objection has teeth, serious people should think about systems that may outrun human cognition. Still, the altitude is suspiciously convenient. Up there, the air is cosmic, the timelines are foggy, and the labs remain the main characters because only they can describe the dragon forming in the cloud.

Down here, in the invoice-and-school-form layer of reality, AI already scores candidates, ranks suppliers, nudges prices, filters customers, summarizes complaints, shapes classrooms, touches healthcare, assists public services, and accelerates war. The current systems do plenty of damage without waking up, reading Nietzsche, and demanding constitutional rights.

Leo’s best word is “disarm.” He does not mean smashing GPUs with a chalice, although some vendor demos test the imagination. He means disarming the assumption that technical power gives anyone the right to govern. Compute grants capability, it does not grant jurisdiction over the future.

That sentence belongs in every boardroom using AI to touch human lives.

The internet also got its little chew toy. AI detectors flagged portions of the encyclical as possibly AI-assisted. Cue the smirking: the Pope warns about AI, maybe with AI. Very funny. Also useful, once the joke grows up. The real question becomes exactly the one the document keeps asking everywhere else: where did assistance end and authority begin? Who reviewed? Who signed? Who owns the words?

That is the corporate AI problem in a cassock. When a consultant drafts strategy with Claude, who owns the advice? When HR scores candidates through a model, who owns the rejected life? When procurement ranks suppliers through a black box, who owns the hidden bias? When a military system accelerates target selection, who owns the dead?

“The model” is cowardice with better latency.

I wrote the full piece on Heliade because this deserves more than a neat little “Pope versus tech bros” headline. The Church has its own ledger. The encyclical knows that. It talks about slavery, abuse, late moral learning, and the ugly fact that institutions often discover the truth after bodies have already paid the invoice.

That may be the most important warning in the whole document: do not make us apologize in 2126 for the digital colonialism we can already see in 2026.

Read the full Heliade piece here: The a[i]theist who needed the Pope

Rome noticed ànd exposed

the bodies under the cloud. That is the embarrassment.

I’ve been watching most of the industry spend three years staring at the same shiny object. VCs, founders, colleagues, journalists, CIOs, and board members who learned the word agentic at the airport lounge in October. Uncle ChatGPT and aunt Claude crashed the innovation party, the corporate music stopped, and everyone is still holding the same lukewarm drink. GenAI: semi-expert of many, master of (almost) none. Meanwhile one of the people who actually built the foundations of modern deep learning has been standing in the corner with a chilled Chardonnay, watching everyone go feral over what he politely keeps calling a useful sub-component.

His name is Yann LeCun. Turing Award winner. Twelve years as Meta’s Chief AI Scientist. The man who invented the architecture that powers every image recognition system on your phone, in every hospital, at every airport. He spent those twelve years saying, on the payroll, that LLMs were a dead end. In November 2025 he left, raised $1.03 billion at a $3.5 billion pre-money valuation in March 2026, and put the headquarters of AMI Labs in Paris. Not London. Not Palo Alto. Paris. Macron stood on a stage at the Grand Palais and lobbied him to stay, the way you usually only see in football transfer windows.

The bet is that the next floor of AI is not a bigger LLM, it is a different building. World models. Systems that understand physics, plan, simulate consequences, and learn from less data. It is the most interesting argument in AI right now, and the most interesting geography play in tech in a decade.

The full piece, with Mistral’s opposite bet, the numbers behind AMI’s round, and why Silicon Valley does not want to hear any of this, is over on Heliade.

Read the full essay on Heliade →

Originally published on Heliade.

Last week a CIO I have known for fifteen years sent me a screenshot on WhatsApp. One developer. One weekend. 5,300 € in coding-agent tokens. He did not write a comment. He did not need to. I knew exactly what the comment would have been.

Cory Doctorow has a word for what is about to happen to enterprise AI: enshittification. First the platform is useful. Then indispensable. Then the value gets quietly rerouted toward the platform owner, once everyone is locked in. SaaS was the rehearsal. Per-employee SaaS spend hit $8,700 in 2025, up 27% in two years, five times market inflation. AI changes the math again, because the bill no longer follows headcount. It follows behavior, ambition, curiosity, panic.

The full piece on Heliade covers why this is the CIO’s second heart attack, what cognitive lock-in actually looks like inside a team, and the FinOps doctrine that has to land before the AI bill becomes the new cloud bill…

Read the full piece on Heliade: https://heliade.net/enshittification-of-ai-cios-second-heart-attack/

A year ago I wrote a piece called Different brain, same meeting invite. A lot of you wrote back. Some of you wrote back very long. A few of you wrote back with what I can only describe as relieved recognition. You asked for more. So here it is. Same wiring. Different bruises.

I burn roughly 75% of my daily energy on the things most of my colleagues do not even notice they are doing. Navigating the noise. Filtering the input. Translating my thinking into linear sentences. Performing socially acceptable enthusiasm. That 75% is the tax. The remaining 25% is what the company is actually paying me for. Now imagine what most of us could do on 80% concentrated on the work that matters and 20% on the rest. (I have done that math more than once. The numbers are devastating.)

The full piece covers open-plan offices, the return-to-office mistake, Monday morning weekend reports, meetings that were never built for brains like mine, the contrarian setting that wants to shake people, the strongholds that keep me functional, and why divergent does not mean superior. Two thousand-something honest words.

Read the full piece on Heliade: https://heliade.net/different-brain-same-open-plan-office/

Originally published on Heliade on 17 May 2026.

Heliade · View from the Edge. Strategy, leadership and the perception from a slightly different brain. The blog is the home. This newsletter is the megaphone.

We spent two years worrying about AI replacing people.

Cute.

The more immediate problem is that every department will start creating its own little digital goblins. Marketing agents. Finance agents. Procurement agents. HR agents. Sales agents. Shadow agents. Personal agents. Agents built by interns. Agents built by consultants. Agents built by people who still name Excel files “final_final_v7_REAL.xlsx”.

That is where this gets funny.

By funny, I obviously mean structurally dangerous, budgetarily stupid, legally foggy, operationally fragile and spiritually exhausting (I have receipts). The usual corporate innovation cocktail, then. Served warm in a reusable cup with a “responsible AI” sticker on it, because apparently irony has procurement approval now.

For a while, we imagined artificial intelligence entering the enterprise through the front door. A big platform. A strategic roadmap. A steering committee. A sober executive sponsor standing in front of a slide called “AI strategic vision”, which is usually where good ideas go to be embalmed. The board would ask its important questions. Procurement would negotiate the license. Legal would squint at the terms. IT would build the control layer. HR would organize a lunch-and-learn where people pretend to understand embeddings while eating sandwiches with too much mayonnaise.

Adorable.

AI will enter the company as mould. Quietly. Everywhere. In corners. In workflows. In browser extensions. In forgotten SaaS tools. In sales teams. In finance. In HR. In procurement. In that one over-caffeinated person who discovered an agent that reconciles invoices and now speaks about it with the dangerous glow of a convert.

Someone in finance will build an agent because the official process still involves six approvals, three systems and a shared mailbox last cleaned during the Obama administration. Someone in sales will build an agent that writes proposals, updates CRM fields and “improves” pipeline notes with the confidence of a LinkedIn influencer after two espressos. Someone in HR will build an agent that summarizes exit interviews, spots morale issues and accidentally creates a litigation starter pack. Someone in procurement will build an agent that compares suppliers, drafts negotiation emails and begins to feel suspiciously like a very cheap McKinsey associate with no parking expenses.

Everyone will be delighted.

For about nine minutes.

Then the CIO will ask who approved the thing. Silence. Then what data it can access. More silence, now with professional eye contact avoidance. Then what actions it can take. At that point someone from the business will say, “Well, technically, it only recommends.”

That sentence should make grown adults reach for a helmet.

“Only recommends” is how many corporate systems already make decisions. The human sees the recommendation, assumes the machine did the thinking, checks nothing, clicks approve, and later explains to risk management that “the process worked as designed” (I have written that sentence, in slides, more than once). Which is one of those phrases that makes you understand why ancient civilizations occasionally threw things into volcanoes.

The numbers are already absurd

The agent problem is already visible. Gartner warned in April 2026 that the average global Fortune 500 enterprise could move from fewer than 15 agents in 2025 to more than 150,000 agents by 2028. One hundred and fifty thousand. Per company. Call that a population explosion with API keys.

I read that number and put the coffee down.

The Wall Street Journal picked up the same scent this month: duplicated agents, cyber exposure, conflicting outputs, governance holes and rising compute costs, all arriving while departments build faster than central IT can track. A familiar corporate rhythm. First everyone wants freedom. Then everyone wants someone else to explain the consequences.

For years, business teams complained that IT was too slow. Often, with reason. IT had become the Department of No, Maybe Later, Please Fill In This Form, We Need Architecture Review (yes, I have run that department). So the business found workarounds. First spreadsheets. Then SaaS. Then automation tools. Then shadow IT. Now shadow AI.

The workaround can read your contracts, write to your customers, browse your knowledge base, call other tools, generate content, trigger workflows and explain its reasoning in soothing English while being fundamentally wrong.

Progress. Apparently.

Shadow AI comes with initiative

Shadow IT was already annoying. Shadow AI comes with initiative.

A spreadsheet sits there and waits to be abused. An AI agent can go looking for trouble. It can call APIs, generate emails, book meetings, update records, search external websites, compare offers, rank candidates, draft supplier complaints at 2.17 in the morning and do all of it with the diplomatic instincts of a wounded badger.

People will trust it too fast because it speaks human. That is the dangerous little magic trick. Software used to look like software: forms, fields, buttons, menus, error messages written by people who clearly lost a bet. Agents are different. They behave like colleagues. They apologize. They summarize. They say “great question.” They explain. They flatter. They remember just enough to feel useful and forget just enough to ruin your week.

The interface is politeness. The backend is authority.

Who owns the consequences?

Every organization now needs to face one brutally simple question: when an AI agent acts, who owns the consequences? The user will point to the manager. The manager will point to the process owner. The process owner will point to IT. IT will point to the vendor. The vendor will point to the model provider. The model provider will point to the documentation nobody read. The consultant who built the prototype will already be on another transformation program with a better day rate and a cleaner badge.

Nobody wants that conversation because the answer is expensive. Governance sounds boring until the machine does something weird with money, personal data, pricing, compliance, supplier terms or a customer relationship. Then governance becomes fascinating. People discover audit trails the way medieval villagers discovered comets: too late, with fear, and usually while pointing.

The job replacement debate has become too narrow anyway. Yes, tasks will disappear. Roles will be compressed. Some people will be exposed as expensive forwarding systems with Outlook access (don’t shoot the messenger). The more immediate organizational shock sits elsewhere. AI agents redistribute action. That is where the power moves. Quietly. Through convenience. Through a workflow improvement. Through a proof of concept. Through a “temporary” workaround that survives longer than the executive who sponsored the official program. That redistribution is also where friction will show its teeth.

Convenience is undefeated.

What goes wrong, in order

The sales agent drafts the offer before pricing has reviewed the latest margin constraints. The procurement agent shortlists suppliers before sustainability has checked the small print. The finance agent flags anomalies using a model nobody can explain at quarter close. The HR agent produces a “sentiment insight” that becomes the ghost in a restructuring meeting. The legal agent summarizes a clause and misses the one sentence that matters, because of course it does. The marketing agent generates a campaign that sounds wonderful until someone realizes the claim is legally optimistic poetry.

Then everyone gathers in a Teams call and performs the ancient ritual of corporate innocence.

“We need to understand what happened.”

No, you need to understand what you unleashed.

The public web is doing the same thing

The public web is moving in the same direction. Cloudflare’s Matthew Prince warned at SXSW 2026 that AI bot traffic could exceed human traffic online by 2027, because agents can visit far more sites than humans while completing tasks. One human shopping for a camera might visit five websites. An agent might visit thousands. Charming. The internet is becoming a place where machines inspect pages written by machines so other machines can decide what humans should see. Very normal. Very healthy. Definitely what Tim Berners-Lee had in mind while not screaming into a pillow.

What happens on the public web will happen inside companies. Internal traffic will become agentic. Internal search will become agentic. Internal knowledge will become agentic. Internal approvals will become agentic. Eventually the company becomes a strange little biosphere where humans supervise agents that talk to other agents relying on documents written by previous agents based on meetings nobody remembers.

A bureaucracy, basically.

Only faster.

The CIO holds the mop

And this is where the CIO’s job becomes both more important and more thankless. The board wants acceleration. The business wants freedom. Legal wants control. Risk wants traceability. Security wants identity management. Finance wants cost discipline. Employees want tools that work before they retire. Vendors want everyone to believe the future can be purchased as a subscription tier called Enterprise Plus Platinum Quantum Something.

The CIO sits in the middle, holding the mop, the map and the fire extinguisher.

Banning agents would be silly, and also useless. People will use them anyway. They always do. Ban something useful in a company and it immediately becomes more popular, like office gossip or decent coffee. The real work is treating agents as a workforce before they become one by accident.

That means identity. Every agent needs an owner. A named one. A human throat to choke, as the old infrastructure people would delicately put it after their third coffee. It means permissions that do not let agents wander through enterprise data like toddlers in a weapons museum. It means scope, expiry dates, escalation rules, logs, cost ceilings and kill switches that work outside PowerPoint.

Especially kill switches.

A real one. “We can always turn it off” whispered by someone who has never tried to decommission anything in a multinational company is the most expensive sentence in the building. If an agent can act, the organization must be able to stop it. Quickly. Cleanly. Without six committees, two vendors and a priest (I have, on at least one occasion, called the priest).

The hardest part is cultural

The cultural layer may be the hardest part. People need to understand when they are using an agent, when they are supervising one, when they are responsible for one, and when the agent is simply wrong with excellent grammar. That last category deserves its own training module, preferably with warning lights.

Because agents will make mistakes in ways humans find difficult to resist. A human junior employee makes a mistake and we question competence. An AI agent makes a mistake and half the room asks whether we prompted it correctly. The machine gets epistemological charity. The intern gets a performance review.

There is something deeply stupid in that, which means it will scale nicely.

The enterprise agent explosion will be messy, political, useful, dangerous, productive, embarrassing and occasionally brilliant. Some agents will save days of work. Some will expose rotten processes. Some will quietly become essential. Some will duplicate other agents. Some will generate beautiful nonsense. Some will be built for one urgent project and then live forever in a forgotten corner of the system, still running every Thursday at 03.00, like a tiny ghost with database access.

That is the work now: finding where these things act, who owns them, what they touch, what happens when they are wrong, and whether the company can still explain itself after the agents have been busy. A company that cannot explain how decisions are made has not transformed. It has become haunted.

And the ghosts are multiplying.

If this hit a nerve, forward it to the CIO, the COO, or the one person in your company who already knows.

Heliade · View from the Edge. Strategy, governance and AI at boardroom altitude

.

You did not see DeployCo trending. There is no Sam Altman thread to argue about. That alone should make you nervous.

What it is

A services company. OpenAI is the majority owner. The lead backer is TPG. The co-leads are Advent, Bain Capital and Brookfield. Around them sits the usual cast: Goldman Sachs, McKinsey, Capgemini, and fifteen others fluent in the ancient corporate dialect of transformation decks and AI readiness scans.

To staff it on day one, OpenAI is acquiring Tomoro, a London AI consulting boutique founded in 2023. Tomoro brings 150 engineers and a client list that includes Mattel, Red Bull, Tesco and Virgin Atlantic. The job title for these new arrivals is “Forward Deployed Engineer.” They embed inside your company, sit next to your people, study how the work actually gets done, then build systems that fit into your workflows.

A Trojan horse. Wooden, beautifully crafted, wheeled right through the gates with the customer’s own hands.

Why they did it

Two reasons. Neither is “to help you.” The first is competitive. Anthropic has been eating OpenAI’s lunch in the enterprise segment for eighteen months. Claude is the model coders reach for. Claude is the model the big banks quietly standardized on. Fidji Simo, OpenAI’s CEO of applications, told staff Anthropic’s gains should be “a wake-up call.”

DeployCo is the answer. If you cannot win on model quality, you compete on getting deeper inside the customer. You stop being a vendor and start being a tenant. The kind that knows where you keep the spare key.

The second reason is structural. AI does not deploy itself. The bottleneck is never the model. The bottleneck is the data fragmented across six systems, the legacy application written in 2003 by a contractor who has long retired to Crete, the works council, the country manager who refuses to give up his spreadsheet. That is the layer where most AI projects die. Aaron Levie, CEO of Box, said it plainly: “Huge opportunity for new service providers.” Translation: whoever owns the integration layer owns the customer.

OpenAI just bought a controlling stake in that layer.

Why it is dangerous

Picture a board meeting. A McKinsey partner walks in with a deck. The deck recommends a phased AI program. Phase one, strategic assessment. Phase two, pilot. Phase three, deployment at scale.

Each phase is staffed by a firm with a financial stake in DeployCo. Each phase deploys a model from a company whose parent owns the firm that staffed the project. Same parent that paid for the deck. Your procurement team will not see this on one slide. They will see four invoices from four brands and assume that means four opinions.

Four invoices. One commercial interest.

The decisions DeployCo helps you make, where AI enters, which jobs become “optimization candidates,” which suppliers get cut, are not technical choices. They are political and economic ones. We used to at least pretend those decisions belonged to the customer, with advisors in advisory roles. DeployCo collapses the distance. The advisor and the vendor are now the same room with the same incentive.

In any other industry, this is a conflict of interest so obvious it would not survive a junior compliance officer. In this one, it gets a press release.

Three questions, if I were on your board

One. When this consultant recommends this model from this vendor, who am I actually paying, and who profits? Get the cap-table answer on paper. If your advisor is also a DeployCo partner, you are not getting advice. You are getting a sales call wearing a tie.

Two. Who owns the workflow knowledge that a Forward Deployed Engineer extracts during deployment? In whose database does that “study” live afterwards? If the answer involves any URL ending in openai.com, your competitive moat just got logged into someone else’s training set.

Three. What is the exit cost?

If DeployCo embeds for two years and then doubles the price, what is your switching cost on day 731? If you cannot answer in dollars, you do not have a vendor relationship. You have a hostage situation dressed up as partnership.

The bigger picture

The headlines will keep being about model benchmarks. GPT-this, Claude-that. The actual game has moved. The interesting question is no longer which model is smartest. The interesting question is who owns the engineer sitting two desks over from your CFO, taking notes. That engineer used to work for you. Soon, they will work for someone whose business card has four logos on the back and one boss.

Think about that, the next time someone tells you AI is “just a tool.”It is a tenant. And the lease just got signed without you reading it.

About this newsletter

This is Heliade · View from the Edge. Short letters from the edge of AI, leadership, and the long road in between. I write the long versions on heliade.net. This is the part that lands in your inbox.

The full DeployCo essay, with all the receipts and the energy/capacity argument I had to cut for length, lives here on Heliade.

If this landed, forward it to one person who needs to read it before their next board meeting. If it did not, tell me why. I read everything.

Danny