Why Incumbents Can't See What's Coming

Pull up any large company's org chart. Go ahead, I'll wait. What you're looking at is a fossil record. Every box, every dotted line, every layer of middle management - it's a solution to a problem that doesn't exist anymore: the problem of translating strategy into execution across hundreds or thousands of people.

That translation layer - all that coordination, approval, handoff, and oversight machinery - used to be the defining competitive advantage of the 20th-century corporation. Now it's the thing most likely to kill it.

---

The Architecture of the Old World

To understand why incumbents are vulnerable, you have to understand why they were built this way. It was brilliant engineering.

Back in 1937, Ronald Coase asked a question that earned him a Nobel Prize: why do firms exist at all? His answer was transaction costs. It's cheaper to coordinate work inside a firm -- through hierarchy, employment contracts, and direct supervision -- than to negotiate every task on the open market (Coase, 1937). The bigger the coordination challenge, the bigger the firm needed to be.

Henry Mintzberg formalized this into organizational theory. He identified the core components of any large organization: the strategic apex that sets direction, the middle line that translates strategy into operations, the operating core that does the work, and the technostructure that standardizes processes (Mintzberg, 1979). Each layer exists for a reason. The strategic apex thinks. The middle line coordinates. The operating core executes. The technostructure makes sure everyone follows the same playbook.

This architecture scaled beautifully for a century. It gave us General Motors, IBM, Procter & Gamble -- companies that could coordinate tens of thousands of workers across continents. The planning-execution divide was the whole point. You needed separate people to plan and separate people to execute because no individual could do both at scale.

Gary Hamel and Michele Zanini estimated that bureaucratic overhead -- all that coordination, supervision, and compliance -- eats more than $3 trillion annually in the U.S. economy alone (Hamel & Zanini, 2020). That price tag was justified when the alternative was chaos. When the only way to get a thousand people pulling in the same direction was to stack them into a pyramid and manage them layer by layer.

So here's the question: what happens when a single person with an AI agent can do what used to require a team of ten?

---

Why the Structure Was Rational

Before you dismiss incumbents as dinosaurs, respect the logic that built them. Their structures solve real problems.

The coordination problem. When your strategy requires 500 people to execute, someone has to break the plan into pieces, assign those pieces, monitor progress, and reassemble the outputs. That's middle management. It exists because humans can't telepathically coordinate at scale. (Wouldn't that be nice, though?)

The quality problem. When execution is distributed across dozens of teams, you need standardized processes, review gates, and approval chains to maintain consistency. That's the technostructure. It exists because variability kills at scale.

The knowledge problem. When tasks are complex, you need specialists -- people who spend years mastering a narrow domain. That requires departments, career ladders, and training programs. Specialization exists because generalists couldn't go deep enough to compete.

The accountability problem. When billions of dollars flow through an organization, you need checks. Procurement reviews. Legal sign-offs. Compliance audits. These exist because unchecked authority leads to fraud, liability, and catastrophic risk.

Every one of these structures was a rational response to the constraints of the pre-AI world. Harvard Business Review research has consistently found that organizational hierarchy enables efficient resource allocation and clear accountability -- when the environment is stable (Guadalupe et al., 2023). And that's exactly the catch. The environment is no longer stable.

---

Why the Structure Is Fatal Now

Here's what changed: AI agents collapsed the planning-execution divide. When planning is doing -- when describing what you want is the same as building it -- the entire coordination layer becomes pure overhead.

Let's walk through each structural element and see what happens when AI enters the picture.

Middle management becomes latency. The middle line exists to translate strategy into tasks, assign those tasks, and monitor completion. But if a founder can describe what they want to an AI agent and get working output in hours, the translation layer is just adding delay. Deloitte found that U.S. employers were advertising 42% fewer middle management positions at the end of 2024 than they did in spring 2022 (Deloitte, 2025). Gartner predicts that by 2026, 20% of organizations will use AI to flatten their structures, eliminating more than half of current middle management positions (Gartner, as cited in Deloitte, 2025). The market is already pricing this in.

Approval chains become competitive disadvantage. Every approval gate that exists to maintain quality also slows things down. When your competitor is shipping features before lunch, your three-week procurement review is protecting irrelevance. A solo operator using AI agents iterates in hours. An enterprise iterates in quarters. Over twelve months, that's a 100-to-1 difference in learning cycles. In a market shaped by iteration speed, that gap is unsurvivable.

Specialization becomes fragmentation. The departmental structure that enabled deep expertise also created silos. Marketing doesn't talk to engineering. Sales doesn't talk to product. Every handoff between departments introduces delay, information loss, and misalignment. Meanwhile, a single operator working with AI agents holds the entire context in their head. No handoffs. No briefs that get misinterpreted. No two-week waiting periods for the design team to pick up a ticket. McKinsey's 2025 State of AI report found that the single biggest driver of AI-generated EBIT impact is redesigning workflows (McKinsey, 2025). The structure itself is the problem.

Compliance becomes paralysis. Risk management in large organizations has metastasized into something that prevents action entirely. Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025, citing unclear business value and inadequate risk controls (Gartner, 2024). The organizational immune system is rejecting them.

---

The Innovator's Dilemma, Revisited

Clayton Christensen described this exact pattern in 1997, long before anyone had heard of a large language model. His insight was that well-managed companies fail precisely because they do everything right (Christensen, 1997). Sounds counterintuitive, right? Stay with me.

The mechanism works like this. Disruptive technologies start out looking inferior by the metrics that matter to an incumbent's best customers. The incumbent's rational response is to ignore them and focus on sustaining innovations -- improvements to existing products that serve existing customers at higher margins. Meanwhile, the disruptive technology improves along its own trajectory, eventually becoming good enough for mainstream use. By the time the incumbent notices, the new entrant has years of compounding learning and the incumbent can't catch up.

AI agents fit this pattern with unsettling precision.

When large enterprises evaluate AI, they compare it against their existing processes using their existing metrics. Can an AI agent match the output quality of a senior marketing team? Not always. Can it navigate our compliance requirements? Not yet. Can it integrate with our legacy systems? Difficult. By every metric that matters to the incumbent's current operation, AI agents look like a toy.

Here's the thing, though. The solo operator isn't measuring on those dimensions. They're measuring on speed, cost, and iteration velocity. And on those dimensions, AI agents are 10x better. The Harvard-BCG study found that consultants using AI completed tasks 25.1% faster at 40% higher quality (Dell'Acqua et al., 2023). GitHub Copilot cut developer task completion time by 55.8% (Peng et al., 2023). These are the kind of performance improvements that Christensen warned make disruptive technologies lethal once they cross the "good enough" threshold.

And they're crossing it now.

---

The Data Shows Incumbents Are Already Falling Behind

The evidence is quantitative and accelerating.

McKinsey's 2025 survey found that only about 6% of organizations have achieved significant EBIT impact from AI. The rest have, in McKinsey's words, "sprinkled AI on top of existing processes instead of rewiring how work gets done" (McKinsey, 2025). That's decoration.

BCG reported that 83% of companies rank innovation as a top-three priority, yet just 3% are "innovation ready" -- down from 20% in 2022 (BCG, 2024). Read that again. The readiness gap is widening. Companies are spending more on innovation and getting worse at it.

Stanford's AI Index found that while 78% of organizations reported using AI in 2024 (up from 55% in 2023), most report cost savings under 10% and revenue gains under 5% per function (Stanford HAI, 2025). Adoption is up. Impact is flat. That's the signature of an organization that has bolted new technology onto old structures.

Meanwhile, Accenture found that companies that embed AI into their core business processes outperform peers by 2.5x in revenue growth (Accenture, 2024). The gap between leaders and laggards is compounding.

---

What Smart Incumbents Could Do (But Probably Won't)

Christensen wasn't fatalistic. He identified a playbook for incumbents who wanted to survive disruption: create an autonomous unit, give it a separate P&L, let it compete with the parent company's own products, and protect it from the organizational antibodies that would otherwise kill it (Christensen, 1997).

Applied to AI agents, this would mean:

Create a small, autonomous team with full authority. An actual operating unit with its own budget, its own hiring authority, and explicit permission to cannibalize existing business lines. This team should operate like a startup -- flat, fast, and accountable to outcomes.

Strip out the coordination layer. Give this team AI agents and let them work without the approval chains, handoffs, and review gates that govern the rest of the organization. Let them ship. Let them fail. Let them iterate at startup speed.

Measure on new metrics. Stop evaluating AI initiatives by whether they match the output quality of a 50-person team. Measure them on iteration speed, cost per outcome, and time from idea to deployment. These are the metrics that will determine who wins in the next decade.

Accept organizational discomfort. The autonomous unit will produce work that's rougher than the main organization's output. It will skip steps that feel essential. It will make the rest of the company uncomfortable. That discomfort means disruption is working. Suppress it and it stops.

Most incumbents won't do this. The organizational incentives work against it. Middle managers won't vote to eliminate their own roles. Department heads won't cede budget to a scrappy team that ignores their processes. The board won't tolerate short-term quality dips in the name of long-term transformation. Hamel and Zanini diagnosed this precisely: bureaucracy is an ideology that privileges control over contribution and compliance over creativity (Hamel & Zanini, 2020). You can't reform an ideology with a memo.

---

What This Means for Small Players

If you're a founder or small operator competing against larger companies, pay attention to what just happened. The structural advantages that large organizations spent decades building -- their coordination capacity, their specialization depth, their process maturity -- have become structural liabilities.

Large companies won't disappear. They have cash reserves, brand recognition, existing customer relationships, and regulatory capture that no solo operator can match. The domains where size was the primary competitive advantage are shrinking fast, though.

Think about what a small team can do now that required enterprise-scale resources five years ago -- once AI is actually connected to the tools where the work happens:

• Ship software products at production quality with a fraction of the engineering headcount.
• Execute marketing campaigns -- scheduling, publishing, analyzing, iterating -- at a volume and consistency that used to require an agency.
• Conduct market research and competitive analysis that used to cost five figures per engagement.
• Handle customer support with AI-assisted triage that matches the coverage of a dedicated support team.
• Build and iterate on financial models, legal documents, and operational playbooks in hours instead of weeks.

Every single one of these capabilities removes a reason for customers to choose the big incumbent over the small challenger. And the gap is widening every month. Bick, Blandin, and Deming (2024) found that generative AI adoption has been as fast as the personal computer. Adoption isn't uniform, though. The operators who go deep -- who wire AI directly into their execution stack instead of toggling between six browser tabs -- are compounding their advantage daily. The bottleneck isn't intelligence anymore. It's the connection between intelligence and action.

The Innovator's Dilemma tells us that disruption doesn't happen all at once. It happens gradually, then suddenly. The incumbents keep reporting record revenues right up until the quarter they don't. By the time the disruption shows up in the financial statements, the structural advantage has already shifted.

We're in the "gradually" phase right now. The operators who are deploying AI -- not just using it to brainstorm, but connecting it to the actual tools that publish, measure, and iterate -- are quietly building the "suddenly."

---

The Org Chart Is the Strategy

There's an old line in management theory: "culture eats strategy for breakfast." Structure eats strategy for breakfast, lunch, and dinner.

You can write all the AI strategy memos you want. You can hire a Chief AI Officer. You can launch an innovation lab. But if your organization is still built around the assumption that planning and execution are separate activities performed by separate people -- if your org chart still has six layers between the person with the idea and the person who ships it -- then your structure will defeat your strategy every time.

The organizations that will thrive in the next decade recognize a simple truth: the org chart is a competitive weapon. And right now, the most powerful org chart in the world is a single operator whose AI agent doesn't just think alongside them -- it acts. It touches the tools. It executes the workflows. One person, operating like a full team, because the gap between deciding and doing has collapsed to zero.

The incumbents can't see this. Their org charts won't let them. And by the time they look up, the solo operator who connected AI to their entire stack will have lapped them.

---

References

Accenture. (2024). Reinventing enterprise models with generative AI. Accenture. https://www.accenture.com/us-en/insights/consulting/gen-ai-reinventing-enterprise-models

Bick, A., Blandin, A., & Deming, D. J. (2024). The rapid adoption of generative AI (NBER Working Paper No. 32966). National Bureau of Economic Research. https://www.nber.org/papers/w32966

Boston Consulting Group. (2024, June 4). 83% of companies rank innovation as a top-three priority, yet just 3% are ready to deliver on those innovation goals [Press release]. https://www.bcg.com/press/4june2024-companies-rank-innovation-as-a-top-three-priority

Christensen, C. M. (1997). The innovator's dilemma: When new technologies cause great firms to fail. Harvard Business School Press.

Coase, R. H. (1937). The nature of the firm. Economica, 4(16), 386-405. https://doi.org/10.1111/j.1468-0335.1937.tb00002.x

Dell'Acqua, F., McFowland, E., Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality (Harvard Business School Working Paper No. 24-013). Harvard Business School. https://www.hbs.edu/ris/Publication%20Files/24-013_d9b45b68-9e74-42d6-a1c6-c72fb70c7571.pdf

Deloitte. (2025). What's the future of management? Deloitte Insights. https://www.deloitte.com/us/en/insights/topics/talent/human-capital-trends/2025/future-of-the-middle-manager.html

Gartner. (2024, July 29). Gartner predicts 30% of generative AI projects will be abandoned after proof of concept by end of 2025 [Press release]. https://www.gartner.com/en/newsroom/press-releases/2024-07-29-gartner-predicts-30-percent-of-generative-ai-projects-will-be-abandoned-after-proof-of-concept-by-end-of-2025

Guadalupe, M., Li, H., & Wulf, J. (2023). Is organizational hierarchy getting in the way of innovation? Harvard Business Review. https://hbr.org/2023/09/is-organizational-hierarchy-getting-in-the-way-of-innovation

Hamel, G., & Zanini, M. (2020). Humanocracy: Creating organizations as amazing as the people inside them. Harvard Business Review Press.

McKinsey & Company. (2025). The state of AI in 2025: How organizations are rewiring to capture value. McKinsey & Company. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value

Mintzberg, H. (1979). The structuring of organizations: A synthesis of the research. Prentice-Hall.

Peng, S., Kalliamvakou, E., Cihon, P., & Demirer, M. (2023). The impact of AI on developer productivity: Evidence from GitHub Copilot. arXiv preprint arXiv:2302.06590. https://arxiv.org/abs/2302.06590

Stanford University Human-Centered Artificial Intelligence. (2025). The AI Index 2025 annual report. Stanford HAI. https://hai.stanford.edu/ai-index/2025-ai-index-report