Cracker Barrel’s CEO is stepping down and being replaced by David Deno. Why?

We may never know the entire rationalization to Julie Masino’s departure but the logo reveal in 2025 is now being narrated as an essential component and significant marketing misstep. I do not believe that to be an accurate narrative.

My guess? It was a Decision Simulation failure. The new branding decisions were from a tight clustering of cognition that rinses and repeats from prior leadership successes rather than reads the (customer) room and executes with realtime telemetry (of alternate decision pathways).

Here’s the breakdown: on August 19, 2025, Cracker Barrel unveiled this ‘modernized’ wordmark. They stripped out ‘Old Timer’ — the overalls-clad figure they call Uncle Herschel. This wasn’t some random whim; it was part of a multi-year strategic transformation plan they’d announced back in May 2024, with roughly $700M on the line.

The backlash was immediate and brutal. The stock lost nearly $100M in market cap in a single session, and even President Trump weighed in. By August 27, they were scrambling to reverse it and put the original logo back. Nearly a year later, CEO Julie Felss Masino is gone, replaced by David Deno.

Now, let’s talk about the team’s cognitive composition. Masino’s resume is a straight line through brand-modernization roles: Starbucks, Godiva, Sprinkles Cupcakes, Fisher-Price at Mattel, and then Taco Bell. She was appointed CEO and a director in 2023 specifically to address the challenges the company faced. That’s a coherent story for sure. But it’s also a cognitive monoculture. Every one of those roles rewards the same instinct: reinvent the visual system, sharpen the value proposition for a younger consumer, and ship the refresh. Pattern driven perfection. Right?

Well the board she was executing with seemed to be aligned with that thesis as well. In 2024, Biglari asked shareholders to vote for him and Milena Alberti-Perez to replace two board directors, Carl Berquist and Meg Crofton. SEC filings even reveal that investor Sardar Biglari warned CEO Julie Felss Masino and her board at least four times the company was in ‘crisis,’ and the transformation plan was ‘obvious folly.’ But the board’s response, in its 2024 proxy materials, was to defend the plan and the slate. That’s not a board with a functioning devil’s advocate; that’s a board with a single, reinforced thesis.

The problem is, what’s missing from the composition is the person who thinks like the Cracker Barrel diner: rural or exurban, older, Southern-inflected, buying the brand precisely because it hasn’t changed. Nobody on the executive team or in the public-facing board bios has an operator résumé serving that customer identity. The team was cognitively long on ‘modernize the brand to grow’ and cognitively short on ‘the brand *is* the moat.’

If you had run the logo decision through an AI Decision Simulation, weighing the cognitive diversity of that room, the signal would have been loud and clear. On the axis that matters (cultural continuity versus brand modernization) the leadership team was heavily clustered on one pole. Every experience they collectively drew from told them modernization was the growth lever. There was no counterweight. The stress-test isn’t ‘will customers like the new logo?’ It’s ‘what happens when this specific team, with this specific composition, has to weigh identity risk against relevancy upside?’ The predicted output is the one that shipped: aggressive modernization, dismissed dissent, and then an identity backlash.

Or they could have asked Tommy Lowe, one of the Co-Founders.

The real lesson here? Cognitive diversity isn’t about demographics, tenure, or even resume variety. It’s about whether anyone in the decision room actually thinks like the customer you’re deciding for, and whether the people who don’t are willing to weight that voice before they ship. Without that, you’re just making guesses, and sometimes, those guesses cost you $100M in a day.

(un)Lonely at the Top · Issue #7

Three industries, two continents, one identical mistake. It was never a strategy. It was cognitive clustering. The AI layoffs of 2024–25 weren’t a strategy failure. They were a cognition failure, the same one we walked through in my Tokenmaxxing article, run against a different budget line. And the tell is that three companies with nothing in common made the identical call, and then made the identical reversal.

The decisions

Between 2023 and 2025, three C-suites in three completely separate industries on three different continents did functionally the same thing. Ford leaned on automated quality inspection while shedding salaried roles. IBM’s AskHR agent absorbed roughly 94% of routine HR interactions, and CEO Arvind Krishna publicly confirmed in 2025 that hundreds of HR staff had been replaced. Commonwealth Bank cut 45 customer-service roles in 2025, tying the redundancies explicitly to a new voice-bot. Different industries, different geos, different stacks, same call - including their reversals.

Ford’s COO Kumar Galhotra told Bloomberg the company had leaned too hard on automated quality systems and the results weren’t good enough; Ford has since rehired, hired, or promoted about 350 veteran specialists over three years and, in 2026, topped J.D. Power’s U.S. Initial Quality Study for mainstream brands, its first time leading the study since 2010.

In February 2026, IBM CHRO Nickle LaMoreaux announced IBM is tripling U.S. entry-level hiring, including for the very software-developer roles “we’re being told AI can do.” And CBA reversed its 45 redundancies in August 2025 and formally called the assessment behind them an error, after call volumes rose rather than fell.

Cognitive composition

Look at who was around each table. Jim Farley, Ford CEO, career marketing and product operator. Arvind Krishna, IBM CEO, career researcher, ex-head of IBM Research. Matt Comyn, CBA CEO, career retail banker with experience in digital transformations. Three very different résumés. Three identical calls.

That’s the tell. When three cognitively distinct leaders arrive at the same questionable decision, the driver isn’t strategy, it’s a shared cognitive frame. In 2024–25 the frame was: adopt AI fast, replace headcount as the visible KPI of adoption, signal cost discipline to the market. Every C-suite was optimizing for the same demonstration.

What was missing from every one of those rooms was a cognition that priced tacit knowledge. The veteran quality engineer who catches a bracket weld failure because he saw the same defect on a line ten years ago. The IBM HR partner who reads the room on a bereavement, a performance appeal, an accommodation request. The CBA rep who de-escalates a scam-loss call at 11pm. That is the 6% AskHR couldn’t route. That is what Ford’s automated inspection missed. That is what CBA’s voicebot could not do.

None of those leaders are stupid. The failure is cognitive clustering: a leadership team made of people who process the same signals the same way, arriving at the same conclusion because no one in the room was wired to price the downside of removing tacit judgment.

What a Decision Simulation would have surfaced

Stress-test the “replace the humans with the agent” call against a cognition weighted toward tacit knowledge and second-order effects, and two questions land in the first pass. What is the cost of the 6% failing? And what is the cost of losing the humans who trained the AI on the other 94%? Both were knowable in 2023. Neither was priced.

The July 2026 news cycle is the evidence the pattern is still running. Volkswagen is weighing up to 100,000 job cuts. Sprout Social cut roughly 20% of its workforce on July 16. KPMG Australia is preparing 1,000+ cuts and reducing partner pay. Meanwhile Orgvue’s survey found 39% of global leaders had cut jobs citing AI, and 55% of those now say the call was wrong. Robert Half data reported by CNBC shows 32% of U.S. hiring managers who cut a role primarily for AI have already rehired for the same or a comparable position. The reversals are documented, on the record, quantified. And yet the cuts continue.

The lesson

Layoffs framed as an “AI strategy” are almost never a strategy. They are the output of a leadership team that shares the same FOMO and cognitive triggers and lacks the diversity around the table to stress-test it before it hits the P&L.

As executives everywhere clamor to replace people with AI, the ones using AI simulations to price the downside before the cut (not after) may be making the most cognitive (and human) choice available to them. And who knows, maybe they can prevent the pain of the boomerang.

By Stuart McClure · (un)Lonely at the Top · Issue #6

On June 30, 2026, Dish filed for bankruptcy and the spectrum Charlie Ergen spent two decades and roughly $35 billion hoarding to build a fourth national wireless network is now being sold, in pieces, to Elon Musk. The Echostar future took a serious hit here but Ergen didn’t lose because he ran out of money or luck. The best decisions are made with full visibility and cognitive diversity, not from the decades old instincts of a poker hand.

The decision

Ergen’s bet was simple and, for a long time, brilliant: accumulate spectrum licenses cheaply, wait, and one day build Dish into America’s fourth facilities-based carrier. Over roughly twenty years he spent about $35 billion on thousands of licenses and earned an industry nickname: the spectrum hoarder.

He answers to no one and controls roughly 91% of EchoStar’s voting power. Before telecom he was a professional gambler, banned from a Lake Tahoe casino in his youth for counting cards. He has described his own strategy in his own words: “I learned to trust my cards. I wasn’t a very good poker player but when a bunch of drunken fools were throwing money around, occasionally I was able to pick up a pot at the end of the day.” That is a strategy for accumulating undervalued assets while others overpay. It is not a strategy for building and operating a network against the clock.

Two clocks caught him. The FCC’s buildout deadlines (cover 70–80% of the population, deploy 24,000 sites) which he kept negotiating to extend, until in May 2025 the Commission accused EchoStar of “warehousing” spectrum. And a $2 billion debt maturity due July 1, 2026. When a planned $23 billion spectrum sale to AT&T slipped, Dish couldn’t make the note. On June 30, Dish DBS and Dish Wireless filed a pre-packaged Chapter 11. Then a week later, on July 6, CEO Hamid Akhavan resigned effective immediately over a “change of strategic direction,” and Ergen took direct control.

Meanwhile the spectrum found a buyer: SpaceX. In September 2025 EchoStar agreed to sell its AWS-4 and H-block licenses to SpaceX for roughly $17 billion to feed a next-generation Starlink direct-to-cell constellation; that November it sold unpaired AWS-3 licenses for about $2.6 billion more in SpaceX stock. The airwaves Ergen couldn’t build on are going to the one operator whose entire cognition is building.

The team’s cognitive composition

Here is what is unusual about this decision room: there was no team to compose. With roughly 91% of the vote, Ergen is the room. The board is an instrument, not a counterweight. And the single cognition running every major call is a genuinely brilliant one, tuned over forty years to a single game: find an undervalued asset, acquire patiently, wait out the reckless, take the pot. It made him a billionaire many times over. It won him spectrum at prices no operator with a build-now mandate would ever have paid.

But that cognition has a structural blind spot, and no one in the room was built to see it. A poker player’s edge is patience: options have value, and time is your friend because the hand resolves when you choose to play it. A wireless network is the opposite. Spectrum is a depreciating, regulated option with a buildout clock bolted to it. Hold it too long and the FCC moves to take it back, the technology curve runs away, and the debt comes due. Ergen kept treating a build problem as an accumulation problem, trusting his cards while the table’s rules changed underneath him.

Now hold that against SpaceX. Musk runs the same governance Ergen does (founder control, no board that says no). Same structure, opposite cognition. Where Ergen’s dominant vector is “acquire and wait,” Musk’s is “build and ship” by putting the asset in orbit and into service before the competition finishes the slide deck. That is why SpaceX is the buyer and Dish is the seller. The lesson is not that founder control is bad; Musk is proof an unchecked founder can also build one of the greatest engineering feats of this or any era.

The lesson is that unchecked founder control amplifies whatever the founder’s cognition already is. It’s a lever, not a virtue. Point it at a blind spot and it magnifies the blind spot.

What a Decision Simulation would have surfaced

Load Ergen’s decision room as cognitive vectors and you don’t get a spread. You get one vector pointed at optionality and patient accumulation, at maximum intensity, with 91% of the voting weight behind it. The stress test writes itself: “Who in this room is modeling the world where the value isn’t in owning the license but in operating the network and where a regulator with a buildout clock and a bondholder with a maturity date force the hand before the asset is built?”

In Ergen’s room the answer is no one, because the room is one man, and that is not how he thinks. A simulation flags that gap in 2018, when it is still a strategy conversation. It surfaced instead as a missed $2 billion maturity in 2026.

The lesson for us all

Ergen may still come out rich. By selling spectrum to SpaceX near the top is itself a masterful trade, exactly the pot-at-the-end move he once described. But the company he set out to build, a fourth national carrier, is in bankruptcy, and its airwaves are moving to a man who builds.

A single cognition, however sharp, wins the game it was tuned for and loses the one it wasn’t. The most dangerous decision room in the world isn’t the one that argues. It’s the one that can’t, because there is only one mind in it and no way to see the blind spot until the clock runs out.

In late 2025, four of the most sophisticated companies on earth told their engineers to use AI as much as possible. By the middle of 2026 those same companies were capping their token budgets in a panic. The reversal wasn’t due to a pricing surprise (although many private frontier models have been steadily increasing their token prices in preparation for their now delayed public IPOs) but rather a cognition failure in controlling spend.

Token Tsunamis: FOMO decision making in 4D

The spend mandate came first. Meta pushed AI tooling so hard it gamified consumption, running an internal leaderboard that staff nicknamed “Claudeonomics” which ranked engineers by usage. The result was a textbook incentive: employees burned an estimated 73.7 trillion tokens in roughly thirty days, and a memo to some 6,000 staff flagged an “exponential increase,” with internal AI costs on track to reach billions of dollars in 2026. One of the best examples of turkeys at the trough with trophies for overindulgence. Now Meta is dismantling the leaderboard and replacing it with a centralized “AI Gateway” to track and ration spend by 2027.

Uber ran the same playbook to the same end. Claude Code adoption across its roughly 5,000 engineers jumped from 32% to 84% in months; heavy users ran $500–$2,000 a month; the company burned through its entire 2026 AI budget by April. Amazon shut down its own internal usage leaderboard in late May 2026 after discovering employees were gaming it. Microsoft pulled back too, cancelling internal AI-coding licenses. The mandate to maximize and the order to cap arrived in the same fiscal year.

Fear of Missing Out infects even the smartest and skilled individuals and teams. Look at who owned these rollouts. They were driven by engineering, platform, and growth leadership, people whose entire professional reflex is adoption, velocity, and capability. More usage equaled more leverage right? Ship faster. The instinct is not wrong; it’s just one axis of thought.

You only turn consumption into a competitive game when nobody in the room is pricing the consumption. “Tokenmaxxing” wasn’t an accident, it was the designed outcome of an incentive built by people optimizing for a number that had no cost attached to it. The cognition that was missing wasn’t intelligence or data; every one of these companies could have forecast the bill. What was missing at the moment of the decision was the unit-economics reflex and the habit of asking “marginal cost per engineer, per token, at full adoption” before the leaderboard goes live, not after. It’s the same structural pattern that took down the OpenAI board: a leadership group clustered tight on one axis, with near-zero coverage on the axis that detonated.

Can you Prevent bad decisioning? Yep, Decision Simulation.

Run the tokenmaxxing mandate through a Decision Simulation. Load each decision-maker’s cognitive profile as a vector. The cluster lights up on capability and adoption and goes dark on cost-consequence forecasting. The stress test writes itself: at 84% adoption, with agentic tools that can burn a thousand times the tokens of a single prompt, who in this room is forecasting next quarter’s burn? If the answer is “no one, because no one around the table thinks in those terms,” the simulation flags it before the incentive is launched, not after a finance memo lands on a Sunday.

The gamification itself is the tell. A team that was modeling the maximum would never have incentivized it.

The Self-Fulfilling Prophecy lesson

This is not an argument against spending on AI, but rather intelligent spending given objective and holistic cognition. The failure was governance. And the swing now underway is from tokenmaxxing to what one might call token turtling, capping everything in a defensive crouch which may be applying the same blindness with the sign flipped: one reflex greedy, one scared, both from a room that still can’t price its own decisions. The waste was never the tokens. It was mandating their consumption from a room that could model capability but not consequence.

The fix isn’t to spend less on AI (although that is clearly coming soon); it’s to put the missing cognition in the room before the decision, so you price the call before the invoice does. Every one of these companies had the data to forecast the bill. None had the cognitive diversity at the table to ask for it and that gap, not the price per token, is what blew up the budget.

By Stuart McClure · (un)Lonely at the Top · Issue #4

A friend shared this book recently “Team of Rivals: The Political Genius of Abraham Lincoln”, and I was enamored and inspired to elevate Lincoln’s story of collaboration mastery to current leadership around the world on this 250th anniversary of America.

The most divided nation in American history was run by a president who staffed his cabinet with the men who had overtly (and covertly) tried to beat him.

That wasn’t magnanimity. It was engineering. Lincoln built the most cognitively diverse leadership team the country had ever seen and the friction between those rivals is exactly why the Union survived. We remember it as a morality tale but truth be told, it was an intentionally purposeful system.

The Team Lincoln built

Lincoln’s cabinet was a collection of men who thought nothing alike and mostly disliked him. William Seward, the polished New York moderate, took the job convinced that Lincoln was a rustic that he could manage from behind the curtain. Salmon Chase, the radical abolitionist from Ohio, considered himself the intellectual superior who rightly should have had the Presidential seat. Edward Bates was a cautious Missouri conservative who distrusted radical change. Edwin Stanton had called Lincoln a “long-armed ape” in his presence and despite this Lincoln made him Secretary of War. Moderate, radical, conservative; Whig and Democrat; free-state and border-state alike. His strategic culture was the opposite of a monoculture, it was assembled by design.

Cognitive Diversity in Action

Watch what the friction produced. In the Trent Affair of 1861, a Union captain seized two Confederate envoys off a British ship and the country howled for a fight such that a war with Britain that would have been suicide. Lincoln used Seward’s diplomatic edge but overrode his appetite for confrontation: “one war at a time,” and the envoys were released. He handed the Treasury to Chase who was openly angling to replace him and whose backers would float the Pomeroy Circular to make him president in 1864. Chase’s financial genius was real: the greenback, the national banking system, the money that funded the war. When Lincoln floated the Emancipation Proclamation, the cabinet split hard: Chase thought it timid, Bates feared upheaval, Seward warned that issuing it amid defeat would read as “our last shriek on the retreat.” Lincoln took Seward’s counsel, waited for Antietam, and issued the proclamation from a position of tremendous strength.

Every one of those outcomes required a view Lincoln did not personally hold, argued by someone who did.

The Lincoln OS (Operating System)

Lincoln’s genius wasn’t tolerance but rather operating discipline, and three rules ran underneath it:

1. Ego was off the table: Lincoln measured himself against the survival of the country, not the comfort of being right, so he could keep a rival who despised him if the rival was useful.

2. Debate as a filter: every major decision was stress-tested against the cabinet’s opposing cognitions (think premortems) before it shipped, which is why so few of them broke.

3. Disagree, then commit (a famous technique by Colin Powell and many other leaders): he let the argument run hot, made the call, and expected total alignment behind it. Diversity in the debate but unity in the execution.

The Division Trap

Now look at today’s political zeitgeist. In a climate of political division, the instinct everywhere, including Washington, boardrooms, and newsrooms, is to reward loyalty and purge dissent and to fill the room with people who already agree and call the agreement “strength”. On the contrary however, it isn’t strength at all. It’s a monoculture, and a monoculture can only see one future which is the exact failure behind every collapse we’ve profiled in this archive. Division and homogeneity look like opposites but they’re actually the same disease. Lincoln’s answer to the most divided nation in history was not to find people who agreed with him. It was to harness the ones who didn’t.

The Final lesson

Here is the part we usually skip but my moral center prohibits in honor of the 250th year of independence and liberty.

Lincoln executed this collaborative mastery by rare intuition, but rare intuition doesn’t scale. You can’t bet a company or a country on a once-in-a-century leader who reads a cabinet by feel. But you can measure it. That is where Wethos becomes the source of record for cognitive diversity: Dynamic Profiling to define your team’s actual cognitive composition (a Team of Rivals, or an echo chamber wearing different titles?), and Decision Simulation to run Lincoln’s debate before the decision instead of after the failure. Cognitive diversity isn’t a virtue you hope you have, it’s a structure that you can see, measure, and manage. Lincoln proved that strategy wins. The only thing that has changed in 160 years is that you no longer have to be Lincoln to run it.

By Stuart McClure · (un)Lonely at the Top · Issue #3

As a former Intel employee, I can tell you that there is an aura in working there. Every week I ran the global CTO office for McAfee under the Intel banner (McAfee sold to Intel in 2010 for $7.78Bn) I felt the burden to live up to their legacy. They had built an incredible brand and mystique that created a near monopoly of chip manufacturing, especially in the US. And they had achieved that feat over decades of disciplined structure and process buildout but the story is in the backstory. But many have not been privy to their stutter steps.

In 2007, Intel’s CEO turned down the chance to build the chip for the original iPhone. He thought the price was too low and it may be the most expensive “no” in the history of technology. But the irony is that Intel kept saying it for fifteen more years. The company’s fall wasn’t bad luck. It was a concretized cognition that could only see one future.

The decision

Paul Otellini said it himself, on his way out the door in 2013: “There was a chip that they were interested in that they wanted to pay a certain price for and not a nickel more, and that price was below our forecasted cost.” In hindsight, he admitted, “the forecasted cost was wrong and the volume was 100x what anyone thought” — “the world would have been a lot different if we’d done it.” Intel protected its margin and ceded mobile.

Then it ceded manufacturing. Around 2015 Intel bet its 10nm process on aggressive multi-patterning with old Deep Ultraviolet (DUV) lithography rather than moving to Extreme Ultraviolet (EUV). The bet failed, 2017 slipped to 2018 slipped to 2019, and TSMC, which had committed to EUV, sailed past into the process lead Intel had held for a generation.

Then it ceded AI. Intel bought Nervana in 2016 to build deep learning chips, then of course shortly after shelved it, bought Habana in late 2019, ran competing roadmaps, and never produced anything that threatened Nvidia, whose GPUs and CUDA ecosystem became the foundation of the entire AI era.

Pat Gelsinger returned in 2021 as the engineer-savior with “five nodes in four years” (5N4Y) and an all-in foundry bet. Instead, after massive capital drains and little to show for it, by December 2024 the board had forced him out. In 2025 the foundry lost $10.3 billion, and the new CEO Lip-Bu Tan cut headcount from roughly 125,000 toward under 100,000 and then to 85,100.

The team’s cognitive composition

For most of this period, Intel’s leadership was grown from the inside. A nearly closed circle of engineers and operators steeped in the x86 franchise and the integrated-device-manufacturer faith: design and fabricate under one roof, defend the margin, and the moat takes care of itself. That cognition wasn’t stupid, it won the PC era and printed money for two decades. But it was a monoculture. Every disruption arrived disguised as a low-margin distraction (e.g. a phone chip too cheap to bother with, a graphics card for gamers, an “accelerator” for a niche called machine learning) and a room that all thought the same way reached the same verdict each time: not our market, not our model, not a threat. And I lived this culture myself.

The tell isn’t any single miss. It’s that the misses rhyme. Otellini protecting margin in 2007 and the leadership defending the integrated model through the 2010s are the same cognition, a decade apart, producing the same blind spot.

What a Decision Simulation would have surfaced

Run any one of these calls through a Decision Simulation and the pattern lights up. Load the room as cognitive vectors and they cluster tightly on margin defense and integration-as-moat, with near-zero coverage on disruption-from-below. The stress test isn’t “is the iPhone chip profitable today?” it is “who in this room is modeling the world where the low-margin entrant owns the volume, the volume funds the process lead, and the process lead laps us?” If the answer is “no one, because no one here thinks like the disruptor,” the simulation flags the blind spot in 2007, when it’s cheap to fix, not in 2024, when it costs the company. Notice who Intel’s board finally hired in 2025: Lip-Bu Tan, a disciplined capital-allocator from outside the x86 priesthood. The missing cognition, installed fifteen years late.

The lesson

To be fair, Gelsinger’s 18A node did ship, and Tan may yet make the foundry work — that verdict is still open. But the open question isn’t the point. A company does not lose mobile, GPUs, and AI in a row by accident. It loses them by composing a leadership team so aligned, so fluent in the worldview that won last time, that the room mistakes consensus for foresight. Intel didn’t miss the future for lack of engineers, capital, or warning. It missed the future because everyone at the table saw it through the same lens: a single lens, however sharp, has a blind spot exactly where the next thing comes from.

Don’t let the market caps fool you. As executives around the globe clamor to embrace AI, those that look to inserting AI simulations to predict future decision success may be the most human choice possible.

Starbucks spent five years getting more efficient and less valuable at the same time. The metrics the leadership optimized went up. The business they were attached to went down. That isn’t a contradiction. It’s what happens when the people at the top can all do the same kind of thinking and none of them does the kind that matters.

The decision

From 2019, under CEO Kevin Johnson, Starbucks leaned hard into digital. Its AI engine, Deep Brew, personalized offers, automated inventory and scheduling, and fed an aggressive push into mobile order-ahead. It worked on its own terms. Rewards members and mobile came to drive more than half of U.S. company-operated sales; order-ahead climbed toward a third of transactions. Laxman Narasimhan, who took over in 2023, doubled down on operational efficiency and “reinvention.”

The numbers they were watching rose. The thing those numbers were supposed to serve (the store) quietly turned into a conveyor belt: mobile-pickup counters jammed, seating disappeared, and baristas were buried under a queue they couldn’t see coming. Seven straight quarters of same-store sales decline followed. Narasimhan was out in just over a year.

The team’s cognitive composition

Look at who set the strategy. Kevin Johnson came from Microsoft and Juniper, a career technologist whose reflex is to instrument a problem and solve it with a platform. His signature Starbucks initiative was, fittingly, an AI engine. Narasimhan spent nearly two decades at McKinsey before consumer-goods operating roles, which was a consultant’s lens, trained to optimize processes and move metrics. Two capable leaders, two different résumés, one shared default: treat the store as a system to be made more efficient.

Howard Schultz built Starbucks on a single non-obvious idea: the “third place” between home and work, where the experience, not the transaction, was the product. Through the decline, nobody at the top carried that lens as their instinct. The tell is almost too obvious: Deep Brew was at one point pitched as AI to counter mobile’s isolating effect, the team had noticed the human experience breaking, and its answer was another algorithm. It’s the same structural pattern as the OpenAI board: a leadership group clustered tight on one axis (efficiency) with near-zero coverage on the axis that actually drove the business.

What a Decision Simulation would have surfaced

Systems thinking, what Oxford’s Mary Johnstone-Louis calls today’s most crucial leadership skill, the ability to see past isolated events to the feedback loops driving outcomes, is exactly the cognition that was missing. Run the mobile-first strategy through a Decision Simulation and the loop lights up: every point of order-ahead throughput you add raises in-store congestion, which degrades the third-place experience, which erodes the very loyalty the rewards dashboard is measuring. Optimizing the part was breaking the whole.

A simulation that stress-tested the strategy against the team’s cognitive composition would have flagged it before the seventh down quarter: no one in this room defaults to asking what speed costs the experience. The dashboard said winning. The system said otherwise.

The lesson

Niccol’s turnaround is the proof by inversion. “Back to Starbucks” — fewer menu items, less digital complexity, more and better-staffed baristas, seating and condiment bars restored, names back on cups. It’s the missing cognition, re-seated at the table. Same-store sales turned positive for the first time in seven quarters and transactions grew. Niccol didn’t out-optimize his predecessors; he saw a different system. That is the entire argument for measuring how a leadership team thinks before it acts: a room that can only optimize will optimize you straight past the thing you actually sell.

On Friday, November 17, 2023, OpenAI’s board removed Sam Altman as CEO, informing him five to ten minutes beforehand, over a Google Meet, as he sat watching the Las Vegas Grand Prix. The public statement was blunt: Altman “was not consistently candid in his communications,” hindering the board’s ability to do its job. They’d lost confidence.

Within days, the decision imploded. By Monday, 745 of OpenAI’s 770 employees had signed a letter threatening to follow Altman to Microsoft unless the board resigned, among them Ilya Sutskever, one of the directors who had just voted to remove him, who publicly apologized for his part. Satya Nadella announced Microsoft would hire Altman and others into a new independent AI unit. By November 22, Altman was reinstated. Of the four directors who ousted him, three were gone from the board within days; only Adam D’Angelo remained.

The root cause? The team’s cognitive composition.

Look at who was in the room: Ilya Sutskever, Adam D’Angelo, Helen Toner, and Tasha McCauley. Sutskever, OpenAI’s chief scientist, was focused on the existential risk of accelerating AI. Toner and McCauley came from the AI-policy and safety world, with backgrounds deeply embedded in the same governance and effective-altruism ecosystem. Adam D’Angelo, the Quora CEO, was the lone operator and the director some observers flagged for a potential conflict, given Quora’s Poe had moved onto OpenAI’s turf with custom chatbots.

Three of the four shared a near-identical lens: AI as a civilizational risk that demands a watchdog board willing to act. The fourth had operator instincts but a contested conflict. Nobody in the room defaulted to operational consequence modeling. No one’s habit was to ask: “What does Microsoft do at hour 24? What do 770 employees do at hour 48? Who runs payroll on Tuesday?”

This wasn’t a values failure. It was a composition failure. The people whose cognition would have caught it: a sitting public-company CEO, a CFO, a comms operator, an investor with skin in the game, had either left the board months earlier or were never seated. Microsoft, OpenAI’s largest backer, didn’t even have a board seat.

Run this call through a decision simulation and the failure mode becomes visible before the vote, not after. Each director’s profile loads as a vector. The board’s collective vector clusters tightly on one axis, governance and safety reasoning, and shows near-zero coverage on the axis that decides everything: “How does a sudden CEO removal propagate through a 770-person company, a multibillion-dollar partner, and a customer base of millions?”

The stress test writes itself. Simulate the decision on a 72-hour horizon and ask whose mental model is doing the predicting. If the answer is “no one’s because no one around the table thinks in those terms,” the simulation flags it before anyone votes. The board didn’t need a different decision. It needed to see, in advance, that it had no one capable of forecasting the response. Five days later, the market told them.

Cognitive diversity at the top is not an HR talking point. It’s the difference between a board that can stress-test its own decisions and one that learns the answer from a push notification on Sunday night.

**Sources**

1. [Removal of Sam Altman from OpenAI — Wikipedia](https://en.wikipedia.org/wiki/Removal_of_Sam_Altman_from_OpenAI) (timeline, “not consistently candid,” board roster, prior board departures)

2. [Hundreds of OpenAI staff tell board to resign or they’ll join Altman at Microsoft — Axios](https://www.axios.com/2023/11/20/openai-staff-letter-board-resign-sam-altman) (745 of 770 letter)

3. [OpenAI brings Sam Altman back as CEO — CNBC](https://www.cnbc.com/2023/11/22/openai-brings-sam-altman-back-as-ceo-days-after-ouster.html) (reinstated Nov 22)

4. [Meet OpenAI’s board of directors that ousted Sam Altman — Fox Business](https://www.foxbusiness.com/technology/meet-openai-board-members-ousted-sam-altman) (board composition)