The Prophets Who Lost


We love simple stories about corporate demise.

The standard story about Kodak is comforting because it casts failure as a moral defect. A lazy monopoly, blind to the future, gets blindsided by scrappy innovators. It turns a complex institutional collapse into a neat cautionary fable about paying attention.

It is also wrong in the way we usually understand it.

Kodak didn’t miss digital photography. Kodak invented the first digital camera and saw the future coming before almost anyone else.

When Steven Sasson built that first digital camera in a Kodak lab in 1975, leadership didn’t just throw it in a closet. Over the next three decades, Kodak invested heavily in digital imaging, built one of the industry’s most significant patent portfolios, and launched early digital photography products and services.

They didn’t lack vision. They saw the future coming miles away. Their problem wasn’t a failure of imagination. It was that their existing business was simply too good.

The Problem with High Margins

Film was not just Kodak’s biggest product line. It was an extraordinary economic engine.

High-margin chemical film funded everything else: the global supply chain, the research labs, the manufacturing infrastructure, the distribution network, and the steady quarterly profits Wall Street expected.

Digital photography wasn’t just a new feature. It was a completely different economic reality.

The engineers looked at digital cameras and saw a technical triumph. The business model looked at them and revealed a hard, uncomfortable truth: every digital camera sold was replacing a high-margin roll of film with a lower-margin piece of consumer electronics.

They weren’t being foolish. Inside the logic of their existing business, protecting film made sense.

They were defending the profit engine that paid everyone’s salary while trying to navigate a market where smartphones, cheaper electronics, and changing consumer behavior were rewriting the economics of photography.

Seeing the future was easy. Making peace with destroying the business that built the company was the hard part.

One of the clearest counter-examples is Fujifilm. They survived not by forcing digital cameras to match film margins, but by realizing their deeper capability was in chemistry, materials science, and precision manufacturing. They adapted by pivoting into healthcare and advanced materials, effectively escaping the photography market’s economics altogether.

A Systemic Pattern

Once you look at history this way, you realize Kodak is not an outlier.

Researchers at Xerox PARC created many of the foundations of modern personal computing. The graphical user interface, Ethernet, object-oriented programming, and the mouse all came out of their labs. Xerox even built and tried to market systems like the Alto and the Star.

Yet Xerox couldn’t capture the value of what it created.

Why?

Because the company’s revenue, sales incentives, and organizational culture were built around high-margin copier leases and moving physical documents. The institution was engineered to optimize one world, leaving Apple and Microsoft to build the next.

IBM is perhaps the most interesting counter-example because they actually managed to reinvent themselves.

Under Lou Gerstner in the 1990s, IBM successfully shifted from a hardware-centric identity toward services and consulting. But IBM didn’t make that transition because executives calmly predicted the future from a position of strength.

They did it because they were losing billions, facing an existential crisis, and backed into a corner. IBM changed only when the cost of protecting yesterday became higher than the risk of destroying it.

The crisis created the permission structure that success had prevented for decades.

Yesterday Is Still Paying Too Well

In executive suites today, enormous amounts of time and energy are spent trying to “predict the future” or sponsor internal innovation labs.

History suggests prediction is rarely the bottleneck.

Kodak built the future. Xerox created the foundations of the future. IBM was forced to rebuild around it.

The harder question for any leader is not whether you can imagine tomorrow.

It is whether you can recognize when the very engine that made you successful has become the thing preventing you from becoming successful again.

The greatest threat to a successful company is rarely that it cannot see the future. It is that yesterday is still paying too well !

The Future Isn’t Built. It’s Unfunded.


Almost every board conversation I sit in on right now follows the exact same pattern.

The strategy decks are full of slides about what comes next. Teams present AI pilots, demonstrate new models, outline partnerships, and describe how the company will become an AI-enabled enterprise. Capital gets allocated to buy tools, hire specialists, and build capabilities.

The assumption behind all of this is simple: If we build enough of the future, we will win the future.

I think that’s only half the story.

The harder question isn’t what you need to build. It’s what you need to stop funding.

The Abundance Trap

Every major technology shift creates the same strategic problem. Something that was once scarce becomes cheap and abundant. When that happens, the source of competitive advantage changes completely.

The mistake most companies make is assuming their old architecture can simply absorb the new technology. It usually can’t.

When electricity hit early 1900s manufacturing, the breakthrough wasn’t simply swapping steam engines for electric motors in multi-story factories built around central drive shafts. As Henry Ford demonstrated at the Highland Park Plant, real advantage required completely redesigning the factory floor around single-story, machine-level motors. Success meant abandoning the entire physical architecture designed for steam.

We saw a similar transition with cloud computing. For years, enterprises talked about cloud adoption while continuing to operate as if physical infrastructure was their center of gravity. They ran isolated pilots, moved workloads selectively, and kept pouring capital into traditional data centers.

Capital One took a fundamentally different path, deciding to exit all eight of their physical data centers and move entirely to AWS.

Looking back at decisions like Capital One’s, a clear diagnostic test emerges for evaluating legacy assets during a technology shift:

Does this legacy asset give us true pricing power, or is it merely protecting a cost position that a competitor can now replicate without owning the asset?

If a technology shock enables a rival or a startup to achieve your hard-won operating efficiency through software or third-party infrastructure rather than capital ownership, your legacy asset is no longer a moat. It is pure overhead.

The Invisible Decisions

Distinguishing between a durable advantage and a stale constraint is where most executive teams struggle, because building something new is always easier than stopping something old.

A well-capitalized company can easily launch an innovation lab, hire specialists, and run pilots. Those actions feel like progress because they are visible and easy to report. The real strategic decisions are usually invisible, and far more painful:

  • Budget reallocation: Which high-margin business unit gets its capital budget cut to fund an unproven initiative?
  • Product sunsetting: Which legacy product line stops receiving engineering maintenance, even if it still generates predictable cash flow today?
  • Incentive overhaul: Which compensation structure gets torn up because it rewards yesterday’s high-margin sales? This is what Microsoft faced under Satya Nadella when unwinding sales incentives built around upfront Windows and Office licenses to force the organization toward cloud consumption metrics.
  • Executive trade-offs: Which successful leader loses headcount and airtime so a new division can grow?

These decisions spark intense internal conflict because yesterday’s success has a loud voice inside the building. The sales team doesn’t want a model that cannibalizes their commission. The engineering team doesn’t want its architecture replaced.

Un-funding the past is also fraught with immense operational risk, because navigating an abundance shift requires solving two completely different problems:

  1. The Diagnostic Problem: Correctly identifying which legacy assets are constraints rather than moats.
  2. The Sequencing Problem: Managing the speed of capital reallocation so you don’t starve your existing P&L before the new model generates real cash flow.

Look at Borders in 2001 for a textbook Diagnostic Failure. When e-commerce emerged, Borders signed an agreement handing its entire online storefront and fulfillment operation to Amazon. At the time, it looked like sensible cost discipline. Amazon had superior logistics, and web retailing yielded lower margins than physical stores. But Borders misjudged its physical superstores as its durable moat and treated the online channel as secondary. By the time they canceled the deal in 2008 to rebuild Borders.com, they had handed Amazon seven years of customer data, transaction history, and digital capability.

Conversely, look at General Electric’s pivot toward GE Digital a decade ago for a classic Sequencing Failure. GE got the diagnosis right. They correctly saw that industrial hardware was being commoditized by software. But diagnosing the problem didn’t solve the sequencing challenge. They pulled capital away from their core industrial footprint faster than their software engine could mature, creating severe organizational strain.

Navigating a technology shift requires getting both right. You have to diagnose what has become obsolete, and you have to sequence the transition without destroying the enterprise. Un-funding the past is not a magic formula for guaranteed victory. It is simply the price of entry for staying in the game.

Reorganizing Around the New Scarcity

AI is forcing this exact moment on us again. Code generation, analysis, synthesis, and routine execution are becoming abundant.

While companies will continue to differentiate through proprietary data, specialized domain workflows, and distribution, the underlying intelligence layers are rapidly commoditizing.

The strategic question isn’t simply: “What AI capabilities should we build?”

The much harder question is: “What parts of our current operating model exist solely because execution was scarce?”

Because those are the exact parts most likely to become your biggest constraints.

A company can drop AI tools onto every desk and still operate exactly the same way it did a decade ago. It can automate individual tasks while preserving the same bloated approval chains, legacy incentive models, and slow decision cycles. That is the modern equivalent of wiring an electric motor to a steam-era belt system. The technology changed, but the system didn’t.

The companies that capture real value from an abundance shock aren’t always the ones that adopt the technology first. They are the ones with the discipline to un-fund what no longer creates an advantage, so they can capitalize on what does.

The question every executive team should be asking right now isn’t: “How do we become an AI company?”

It’s: “What are we still funding simply because it made us successful yesterday?”

That’s where the real transformation begins.

In the next post, I’ll examine one of the most persistent myths in enterprise strategy: the idea that the giant companies that got displaced simply failed to see the future coming. The historical record suggests something much more interesting. Many of the giants that lost actually saw the future earlier than everyone else. They just couldn’t reorganize around it.

The Functional Enterprise


What Survives When Expertise Becomes Free

AI is making expertise abundant, but modern corporations were built around the opposite assumption: that expertise was scarce and had to be concentrated. That assumption shaped the modern company because every major technology revolution changes more than the tools people use. It changes the way organizations decide, coordinate, and create value.

The Industrial Revolution did not just create factories. It created a management system for coordinating thousands of workers and machines, and hierarchy emerged because industrial scale required a new way of organizing people, processes, and decisions. As companies became larger and more complex, a different constraint emerged. The problem was no longer simply coordinating labor, but finding, distributing, and applying specialized knowledge. The answer was the functional enterprise: Sales, Finance, Marketing, HR, Legal, and Technology.

These departments feel like permanent features of business today, but they were historical responses to a specific problem. Expertise was difficult to create and even harder to distribute, so companies built functions around where specialized knowledge lived. The org chart reflected where the brains were parked. Financial expertise lived in Finance, customer understanding in Marketing, technical expertise in Technology, and legal judgment in Legal.

For more than a century, this architecture worked because expertise, authority, and accountability were bundled together. The people producing the analysis were usually the people trusted to exercise judgment.

AI breaks that relationship.

For the first time, expert-level analysis can be generated outside the functions that historically owned it. A product manager can evaluate financial scenarios, a salesperson can build sophisticated proposals, and an operations leader can diagnose complex problems without assembling a room full of specialists.

The important shift is not that expertise disappears. It is that expertise is no longer confined to the place where the organization expects to find it.

This creates a new management problem. When everyone can generate expert-level analysis, the production of expertise is no longer the scarce resource. Judgment is. Anyone can manufacture an answer. The harder question is deciding which answers deserve to be trusted.

This is where much of the AI conversation stops too early. The question is not whether machines can produce answers. It is how organizations preserve judgment when the production of insight is no longer limited to the experts who historically controlled it.

That challenge extends beyond today’s organization. It also raises a question about tomorrow’s leaders. For generations, institutional judgment was built through the manual work of junior professionals: building spreadsheets, reviewing contracts, writing code, and learning from mistakes. If AI increasingly performs the production of analysis, the traditional apprenticeship model that produced future CFOs, General Counsels, and CTOs begins to disappear. How organizations develop judgment in an era where fewer people learn by doing may become one of the defining management questions of the AI age. It deserves a deeper discussion than this post allows, and I’ll return to it in a future post.

Furthermore, while the generation of analysis can be decentralized, legal liability and regulatory accountability cannot. A product manager might run an autonomous financial scenario, but the CFO still signs the Sarbanes-Oxley certification. A salesperson might use an agent to draft a contract, but the General Counsel still carries the ultimate fiduciary responsibility.

Functions were never valuable simply because they produced analysis. They mattered because they created trusted places where judgment could live, and where ultimate accountability could be held. Finance was never valuable because spreadsheets were difficult to create. It was valuable because someone had to decide where capital should go. Legal was never valuable because contracts were hard to review. It was valuable because someone had to decide which risks the company was willing to accept. Technology was never valuable because writing software was impossible. It was valuable because someone had to decide which systems could safely run the business.

AI does not remove those responsibilities. It changes where the value sits.

The risk is that functions confuse authority with judgment, retaining approval rights while gradually losing the deep understanding required to make good decisions. Faced with that shift, the natural instinct will be to defend traditional boundaries, budgets, and approval rights. That response protects the function, but it does not strengthen the enterprise. It simply turns the function into a bottleneck.

A Finance organization that simply approves AI-generated analysis becomes a rubber stamp. A Legal organization that blindly accepts AI-generated contracts becomes a risk multiplier. A Technology organization that governs intelligent systems without understanding how they fail becomes a governance layer without technical depth.

The functions that matter in the future will not be the ones that produce every answer, nor will they be the ones that use policy to slow execution. They will be the ones that stay closest to where judgment fails.

The answer to the trust question is not the function with the most authority. It is the function with the strongest understanding of the consequences when judgment fails. That understanding cannot be preserved through approval rights alone. It comes from staying close to failure.

Crucially, this failure is rarely a single catastrophic event. In an AI-driven enterprise, the greater threat is silent drift: the slow, almost invisible erosion of quality across thousands of automated micro-decisions.

Preventing that drift requires an active connection to reality. A Finance team maintains its judgment by understanding why forecasts miss. A Legal team maintains its judgment by studying where automated reviews overlooked nuance rather than waiting until a public dispute exposes the gap. A Technology team maintains its judgment by understanding how intelligent systems behave under real-world conditions rather than controlled demonstrations.

The old enterprise was designed to move expertise. Work flowed from Sales to Operations, from Operations to Finance, and from Finance to Legal because knowledge lived inside the function. That logic begins to break down once expertise becomes widely available. The challenge is no longer moving work to where the knowledge sits. It is making sure the organization learns faster than the complexity it creates.

That requires a different operating rhythm. Finance cannot wait for the annual planning cycle to discover whether its assumptions were wrong. Legal cannot learn only after litigation exposes weaknesses in contract review. Every function has to shorten the distance between prediction and outcome, not because every consequence arrives quickly, but because an organization that learns slowly ends up governing tomorrow’s decisions with yesterday’s assumptions.

This is why functions will not disappear. They will become more important, but for a different reason.

For more than a century, companies organized themselves around where expertise lived. The next generation of companies will organize themselves around where judgment is created, challenged, and continuously refined.

That is a different kind of institution.

Most executives think they are introducing AI into a twentieth-century corporation.

They are discovering that the corporation itself is the legacy system.