The Future of Companies: Same AI. Opposite Futures

Picture two ordinary weekday mornings, eight months apart. On the first, roughly four thousand people at Block open their laptops to a meeting invite that turns out to be their last. Forty percent of the company, gone before the coffee was cold. Wall Street delivers its verdict within the hour: the stock jumps. 

On the second, a call-center worker at IKEA logs in expecting another shift spent chasing late deliveries and is told instead that she is being retrained to design people’s living rooms over video. Her job is not being deleted. It is being promoted. 

The same machine sits behind both mornings: software that can now do the routine thinking people used to do for one another. The contrast is not an anomaly. It is the story. 

Two companies, handed the identical capability, built opposite organizations out of it. They cannot both be the future of the company, and that is precisely the point. The machine did not choose.

Sources: Forbes 2026 · CNN 2026 · PYMNTS 2026 · Ingka Group 2026 

The people running each company chose, and the choice gave away what they had believed about their people all along. We have spent two years asking what AI will do to companies. The more revealing question, now visible in real org charts instead of predictions, is what AI exposes about them. 

The morning Block erased itself 

First, who Block is, because the name is younger than the company. Block is Jack Dorsey’s other company. Dorsey co-founded Twitter; Block is the business once called Square, the one behind the little white card readers in coffee shops and the Cash App on half the phones in America. In its last fiscal year, it turned a gross profit above ten billion dollars and raised its guidance. It was not in trouble. 

On 27 February 2026, Dorsey cut headcount from over ten thousand to under six thousand in a single day, about forty percent, and told everyone the cuts were not weakness but a bet: a smaller team, using the tools Block was building, could do more. The market loved it. The shares rose about twenty-two percent. 

Those “tools we are building” have a name, and it matters, because a strategy you cannot name is just a press release. The tool is goose, Block’s own open-source AI agent, released in early 2025. Built in Rust, it runs wherever a company’s data already lives, a deliberate choice for a firm that moves money, and it plugs into more than a dozen large language models through the Model Context Protocol, the open standard Block developed with Anthropic. Inside Block, goose already drafts code, sorts of sales leads, manages content, and onboards new hires; project managers say it has cut their admin time by roughly three quarters. Layered on top is the part that actually reshapes the org chart: a “company world model,” an always-current, machine-readable picture of the entire business, assembled automatically from the artifacts of remote work, every decision, ticket, design, and commit. Dorsey and Botha call the ambition a “mini-AGI.” 

Here is the logic that follows. In a normal company, a manager’s real job is to hold that picture in their head and relay it up and down the chain: what is being built, what is blocked, who needs what. If the model now holds that picture continuously, the relay layer has nothing left to relay. Block collapses the org into three kinds of people. 

Everything in between, the middle-management layer whose whole purpose was to carry messages up and down, disappears. 

We tried this in 2014 

Before you applaud, a confession of my own. In 2014 I flew to Las Vegas for Zappos Culture Camp, the immersive program the shoe retailer ran to teach outsiders its famously joyful culture, the one built on a first core value of “deliver WOW through service.” I came home, a believer in most of it: the warmth, the obsessive care for customers, the sheer nerve of the experiment. But one part unsettled me. The people evangelizing Holacracy, the manager-free operating system Zappos was rolling out, did not behave like guides. They behaved like missionaries. They were dogmatic. They did not really want to discuss whether it worked. They wanted you to convert. 

The numbers arrived later, and they were not kind. Zappos began dismantling its hierarchy in 2013: no managers, no job titles, no org chart, just self-organizing “circles.” In March 2015, CEO Tony Hsieh sent a 4,700-word memo giving staff an ultimatum, commit fully or take a buyout. Eighteen percent walked, around 210 of 1,500 people. The company reorganized itself into 460 circles. Then it stalled. Decisions froze; as one employee put it, you were empowered to act, but nobody wanted to be first. By 2017 and 2018 the managers had quietly returned, the whole thing rebranded a “market-based ecosystem.” Medium, founded by another Twitter co-founder, Ev Williams, tried Holacracy too, and abandoned it. 

So, strip the AI away and Dorsey’s idea is old. Remove the managers, flatten the chart, and trust people to coordinate themselves. Dorsey and Botha are honest enough to say so: their own essay lists Holacracy, Spotify’s squads, and Valve’s flat structure among the experiments that eventually reverted to hierarchy, because the coordination work is real and someone, or something, has to do it. This is the whole distinction, and the whole gambling. Zappos removed the managers and handed the coordination back to humans running a rulebook by hand. Dorsey is betting that a machine does it instead. That single variable is genuinely new. Whether it is enough is the open question of the decade. “Parts of it will likely break before they work” is a more careful sentence than Hsieh’s “rip the band-aid,” but it is the same leap of faith, taken eight years later. 

The billion hiding in the bot’s failures 

Now walk into an IKEA on a Saturday. You know the choreography: the meatballs, the arrows painted on the floor, the flat-pack labyrinth you cannot escape without buying a plant you did not come for. Somewhere in the bedroom section, a couple stands in front of a wall of shelving, phones out, trying to work out whether the KALLAX fits the alcove back home. They do not want a delivery date. They want someone to tell them what works. 

That couple is the reason IKEA’s largest franchisee, Ingka, grew a billion-euro business out of the chatbot’s shortcomings. 

The bot is called Billie, after the Billy bookcase. By 2026 it was resolving about fifty-seven percent of customer enquiries, up from forty-seven at launch, saving an estimated thirteen million euros along the way. The next expected step was to bank the savings and quietly release the call center staff Billie had displaced. Ingka did the opposite. It studied half of enquiries Billie could not resolve and found they were not about broken shelves or missing screws. They were that couple in the bedroom aisle: people asking for help designing a room. So Ingka retrained roughly eight thousand five hundred call-center workers into remote interior-design consultants, paid video sessions, real consultative work, and grew that channel into one worth about 1.3 billion euros, aiming for ten percent of total sales by 2028. Ingka’s CEO, Jesper Brodin, was blunt about why it was worth it: most of the old jobs, he said, were “soul crushing.” Retraining was an opportunity, not the cost. 

None of this makes IKEA a saint, and the tidy fable is worth resisting. In 2026, the group also cut around 1,650 corporate and administrative roles, blaming soft demand and US tariffs, even as it told the reskilling story elsewhere. And the celebrated revenue figure needs a second look: that 1.3 billion euros is the entire remote-selling channel; not money you can pin solely on the retrained workers. Still, this is an experiment worth studying closely, precisely because it does something rarer than a layoff. It asks what the machine’s failures reveal and then builds on to the answer. 

Your org chart is a confession 

Put the two companies' side by side and the uncomfortable truth surfaces. Identical capability produced opposite organizations. The capability is not what produced them. The technology was constant. What differs is what each leadership team already believed about people: overhead to be removed, or capability to be redirected. 

This is the quiet law of AI in the enterprise: it amplifies whatever culture it lands in. A company that already saw its people as a cost will use AI to prove itself right. A company that saw its people as untapped capability will use AI to find out how much. The machine is a mirror. Same machine, opposite reflection. 

Same machine, opposite org chart

And there is a thumb on the scale worth naming out loud. A Harvard Business Review analysis in January, drawn on a survey of more than a thousand executives, found that companies are cutting jobs because of AI’s potential, not its performance: the layoffs are pre-emptive, not earned by results. Pair that with a market that rewards the shape of subtraction with a twenty-two percent pop on the day, and the future of the company starts getting written by what earns applause at the announcement, not by what works two years later. The cut is legible and instant. The reskill is slow and quiet. Capital can see one and not the other. 

The two companies are not even binaries. They are poles on a dial, and the experiments in between show the choice is a setting, not a switch. 

What to actually do 

Whose call is this? Not the CTO’s. Technology is now a commodity: every company in your sector can buy the same model by Friday. What cannot be bought is the decision about what your people are for, and that decision sits with leadership, in the open, on the record. For example, Klarna’s reversal is proof that getting it wrong is expensive: it replaced the work of about 700 agents with an OpenAI-powered bot, then rehired humans after the CEO admitted the all-AI approach produced “lower quality” service.

Here is the work that the decision actually requires. 

 Same machine, opposite org charts 

Question to you: When you look at your people through this new machine, do you see a cost to remove, or a capability you have never fully used? 

Ten years from now, AI won't explain why companies won or lost. Their org charts will! 

Tell me which you see, and why. Hit reply. I read every response, and the sharpest answers will shape where this series goes next 😉. 

Sources 

[1] Forbes (April 2026), Josipa Majic: Jack Dorsey Bets 4,000 Jobs That AI Can Replace The Org Chart. Block cut headcount from over 10,000 to under 6,000 on 27 February 2026, roughly 40%, from a position of strength (FY2025 gross profit $10.36B, up 17%, guidance raised); shares rose about 22%. CNN (26 February 2026) reported the cuts. 

[2] Jack Dorsey and Roelof Botha (31 March 2026): From Hierarchy to Intelligence, published on Block and Sequoia. The Roman-army framing, the “company world model” and “mini-AGI,” the three-role structure, and the list of prior flat-org attempts that reverted to hierarchy. block.xyz/inside/from-hierarchy-to-intelligence and sequoiacap.com/article/from-hierarchy-to-intelligence. 

[3] Block Open Source (January 2025) and VentureBeat (December 2025): codename goose, Block’s open-source AI agent framework, built in Rust, running where data lives, connecting to 15+ LLM providers via the Model Context Protocol developed with Anthropic; internal use for sales analysis, content, and onboarding, with project managers reporting about 75% less admin time. 

[4] Fortune (April 2026): coverage of the Dorsey and Botha essay, middle management’s forecast “extinction,” and the comparison with Amazon’s 14,000 corporate cuts to remove organizational layers. 

[5] Quartz, Time, Fortune, HR Dive (2015 to 2020): Zappos and Holacracy. The 2013 rollout, the March 2015 ultimatum memo, the 18% buyout (about 210 of 1,500), 460 circles, the quiet reversal by 2017 to 2018, and Medium’s abandonment of Holacracy. 

[6] Zappos Insights: Culture Camp, the company’s immersive culture-training programme in Las Vegas, and Zappos’s first core value, Deliver WOW Through Service. Author’s own attendance, 2014. 

[7] PYMNTS (May 2026) and Ingka Group newsroom: IKEA Turned 8,500 Call Agents Into Design Consultants. Billie’s resolution rate rising from 47% to 57%, an estimated €13M saved, roughly 8,500 workers reskilled into remote interior-design consultants, a remote channel worth about €1.3B with a 10%-of-sales target for 2028, and Brodin’s “soul crushing” remark. 

[8] Employer Branding News (June 2026) and Reuters: IKEA’s roughly 1,650 corporate and administrative role cuts in 2026 (soft demand, US tariffs), and the caveat that €1.3B represents the entire remote-selling channel, not revenue attributable solely to reskilled workers. 

[9] Bloomberg (May 2025), Entrepreneur, and Forbes: Klarna replaced the workload of about 700 customer-service agents with an OpenAI-powered assistant (about 2.3M chats, 75% of volume), then reversed course as CEO Sebastian Siemiatkowski acknowledged “lower quality,” rehiring humans in a flexible, remote setup. Gartner has predicted that by 2027 half of companies that cut service staff for AI will rehire. 

[10] Forbes (April 2026), Jodie Cook: reporting that the telehealth startup Medvi was on roughly $1.8B in projected annual revenue with two employees. Figure as reported; treat as directional. 

[11] Harvard Business Review (29 January 2026), Thomas H. Davenport and Laks Srinivasan: Companies Are Laying Off Workers Because of AI’s Potential, Not Its Performance, based on a December 2025 survey of 1,006 global executives. hbr.org