Three Questions to Know If Your Company Will Survive 2026

At every inflection point in history, there are a handful of screening questions that separate the companies that make it from those that don't.

In June 2026, McKinsey published four reports within the same month. They cover four seemingly distinct domains—geopolitical scenario planning, the operating truths of AI-native companies, the industrialization of intelligence via AI, and the redefinition of human roles in tech transformation. But when read together, they point to a single underlying thesis: the market is screening companies on a new dimension—not on technological sophistication, but on the maturity of organizational judgment [1][2][3][4].

Assessing that maturity requires three questions.

I. Can Your Organization Spot a Black Jellyfish?

In a June 2026 report titled The Art, Science, and Technology of Geopolitical Scenario Planning, McKinsey's Geopolitics Practice introduced a new category of risk: the Black Jellyfish [1]. The term draws its name from a real event. In 1999, a massive jellyfish bloom clogged the cooling intake of a power plant in the Philippines, triggering a major blackout. Operators and authorities initially misattributed the outage to other factors—including a coup attempt—before engineers identified the real cause [1].

The Black Jellyfish is defined as: "events for which the triggers are known but the ripple effects are poorly understood and can escalate rapidly" [1]. You know jellyfish are in the water, but you cannot predict when they will surge into the cooling intake, how completely they will clog it, or how far the cascading effects will spread.

When you map this framework onto today's business environment, the scenes are instantly recognizable to any management team:

You know trade tensions are escalating (jellyfish known), but cannot predict which product categories the next tariff will hit or at what rate (clog location unknown);
You know global supply chains are reorganizing, but you don't know which second-tier supplier's second-tier supplier will halt production first (cascading effect unknown);
You know AI regulation is tightening, but cannot anticipate when compliance thresholds will shift or in what form (escalation speed unknown).

McKinsey's concurrent global survey (February 2026, 202 respondents, 82% from C-suite and C-suite-1 level) reveals two troubling data points [1]:

First, fewer than one-third of respondents consider their organization's management of geopolitical risk "mature." Second, while 53% say scenario planning and other strategic foresight methods would "significantly strengthen" organizational resilience, fewer than 30% report actually using such tools to guide decision-making. Only 5% have conducted simulations—the lowest adoption rate among all foresight methods [1].

Even more revealing: AI-assisted geopolitical intelligence and early warning emerged as the biggest capability gap—cited by 45% of respondents. It was followed by "integration of geopolitics into core business decisions" (36%) and "scenario-based planning and foresight" (33%) [1].

What does this mean? It means the vast majority of organizations face Black Jellyfish with no early warning system—not because risk data is unavailable, but because there is no organizational habit of systematically thinking through unpredictable cascading effects.

Chevron CEO Michael Wirth makes the point in a passage quoted in the McKinsey report [1]: "Even if you run drills, you are never going to anticipate the black swan. But … you start to build muscle memory for asking the right questions and for thinking about the collateral impacts that are very difficult to identify in the heat of the moment."

The phrase "muscle memory" deserves close attention. It implies that the capacity to navigate uncertainty is not a one-time construction project. It requires repeated practice embedded in the organization's daily rhythm—a form of embodied organizational knowledge.

The report recommends a full toolkit of five foresight instruments: horizon scanning, scenario planning, contingency planning, simulations, and tabletop exercises [1]. Their value lies not in "prediction" but in a principle the report returns to repeatedly: "Plans are useless, but planning is indispensable." What this really means is that the process of planning is itself capability building.

II. Is Your Knowledge Flowing Water, or Stagnant?

Spotting the Black Jellyfish is only the first step. The real question comes next: When the signal arrives, what level of information can your organization mobilize to make a decision?

McKinsey's Technology Practice published a second report in the same month—The Seven Operating Truths of AI-Native Companies—based on in-depth interviews with 15 AI-intensive companies. It surfaces one of the most underappreciated management propositions in recent years: Knowledge Hygiene [2].

The report states its core thesis in two judgments [2]:

"Your model isn't the bottleneck—accessing your tribal knowledge is."

"The ceiling on your AI is set by your knowledge hygiene, not your model choice."

What this means in practice: When your AI agent cannot answer a question, it may not be the model's fault. The answer may simply never have been written down, or it may exist somewhere the model cannot reach.

The operations director at an energy tech platform puts it more bluntly [2]: "It isn't an AI problem—it's a knowledge management problem. AI just makes it visible."

The second half of this sentence is worth dwelling on. The knowledge management problem was always there, but it was buried, invisible, non-lethal—until AI came along and pressure-tested the organization's knowledge layer. In this sense, AI isn't creating a new problem. It's conducting a stress test—and exposing fractures that were always there.

The report catalogues the classic symptoms of poor knowledge hygiene [2]. Almost every organization will recognize itself in at least one:

Meetings go unrecorded and untranscribed; critical decisions live only in participants' notes.
Expertise exists only in senior employees' heads; when they take vacation, the team is paralyzed.
The latest version of a key document is on someone's local drive, while the shared folder contains last year's outdated version.
An AI agent confidently retrieves contradictory information—eroding organizational trust one interaction at a time. As the COO of a Series D fintech venture warns: "An agent doesn't know what is the latest source of truth and what is an outdated document from a year ago." [2]

A digital health company's CTO admits to learning this the hard way [2]: "If I could change one thing, I'd invest earlier in structuring our content. Fragmented data slows down flow and frustrates teams."

The contrast with organizations that get knowledge hygiene right is equally vivid [2]:

At a seed-stage AI company, every sales call is automatically recorded, transcribed, and routed to a shared knowledge layer. The CEO can ask the system, "Where are we with a specific client?"—and get instant deal context pulled from months of accumulated conversations. The feedback loop becomes a competitive advantage. At an energy tech platform, a knowledge agent indexes code repositories, Notion pages, and Slack conversations—new hires achieve full productivity in days, not weeks [2].

Notably, the agtech CEO interviewed by McKinsey challenges the "single source of truth" orthodoxy [2]: "A lot of people get wrapped around the axle of 'you need a single source of truth.' But the thing that makes data useful is that humans are touching it and updating it all the time." Instead of imposing uniformity, the company builds lightweight connectors that make all data—wherever it lives—queryable by AI.

This aligns directly with Suman Thareja's argument in Rewired's second edition [3]: "A typical learning journey doesn't cut it anymore. Learning has to happen in the flow of work—cross-functional, hands on, and together." The essence of knowledge hygiene is not a one-time cleanup. It is embedding knowledge creation, capture, querying, and iteration into the natural flow of everyday work.

At this point, a dangerous feedback loop emerges between question one and question two: Black Jellyfish proliferate (the density of external shocks increases) → organizations need higher-quality information to make decisions (cognitive load of decision-making rises) → but knowledge acquisition, organization, and refresh fail to keep pace (the knowledge layer rots faster than the environment shifts) → decision quality declines → the next Black Jellyfish catches you even more unprepared. Spotting the jellyfish is the starting point. The quality of the knowledge base you can mobilize to respond determines whether you get consumed by it—or navigate around it.

III. How Much Autonomy Do You Trust Your AI With?

Snowflake CEO Sridhar Ramaswamy, speaking on the June 2026 McKinsey Podcast, delivered what may be the most condensed assessment of the 2026 AI competitive landscape [4]:

"The real business constraint is no longer code, but judgment."

The power of this statement lies in how it reverses three years of anxiety-driven narrative. The dominant narrative has been: "AI will replace me; my skills will depreciate." Sridhar's point is that the direction is wrong. Code has been replaced by AI—but code is no longer the constraint. The new constraint is judgment: when to trust AI output, when to override it, when to delegate, and when to pull back.

Why? Because AI has driven execution cost to near zero. It can generate 10, 100, or 1,000 options in an instant. But the cost of choosing which option to pursue has not decreased—it has increased to an unprecedented level.

The Seven Operating Truths of AI-Native Companies devotes an entire principle to this challenge—Trust Precedes Autonomy, the fifth operating truth [2].

The research finds that the most effective AI-deployment teams are not the ones that delegated earliest. They are the ones that followed a disciplined trust ladder [2]:

Human-in-the-loop: All AI outputs are reviewed by humans, who hold final approval authority.
Human-on-the-loop: AI executes routine tasks autonomously, but humans monitor the process and can intervene at any time.
Human-out-of-the-loop-for-routine: AI handles well-validated routine scenarios fully autonomously, alerting humans only for exceptions.

Each "promotion" to the next level depends on accumulated error-rate data from the previous level. This is fundamentally a trust decision—not a technology decision. Who sets the cadence of delegation? Where is the error-rate threshold? When should human oversight be reinstated? The ability to make these judgments correctly is what constitutes genuine organizational competitiveness [2].

The report uses the term "slow automation" to describe the teams that performed best [2]—having humans execute processes manually until pain points crystallized and trust data accumulated, then introducing AI to solve the specific, well-understood pain. The process looks "slow" but avoids an exponentially larger cost: scaling AI too fast, suffering a trust crisis, and tearing down the whole system to rebuild.

One case from the research involves an AI-native go-to-market startup on Salesforce's ecosystem [2]. Using AI agents, the company's transaction volume increased from 50 to 3,000 per month—without adding headcount. The AI handled screening, matching, and initial outreach; the human sales team was redeployed into high-touch lifecycle management. This is Trust Precedes Autonomy manifested at the organizational division-of-labor level. The ability to decide which processes AI can handle autonomously is itself the most critical judgment an organization needs to cultivate.

Sandra Durth, writing in Rewired's second edition, provides an even more fundamental litmus test for judgment [3]:

"Are you using AI to unlock short-term productivity—or to build durable competitive advantage?"

This question distinguishes two fundamentally different AI strategies [3]:

Efficiency strategy: Use AI to do what you already do, at lower cost. The results are immediate and measurable—but they are also replicable. Your competitors are doing the same thing.
Capability strategy: Use AI to do what was previously impossible—gather intelligence you could never access, make predictions you could never make, deliver services you could never deliver.

The first option makes next quarter's numbers look better. The second determines whether the organization will exist in five years. The second operating truth from The Seven Operating Truths—"Build what makes you distinctive" [2]—forms an inside-out relationship with Durth's question: only after clarifying "what am I building toward" can you answer "what should I build versus buy."

The Three Lines of Judgment

When threaded together, the three questions form a layered defense of judgment:

Line	Core Question	Cognitive Anchor	Key Data Point
First	Spotting the Black Jellyfish	External environment sensing	Only 5% run simulations [1]
Second	Knowledge Hygiene	Internal information foundation	AI only makes old problems visible [2]
Third	Trust Precedes Autonomy	Organizational division of trust	Judgment, not code, is the new constraint [4]

The three lines form a progression: External perception (Black Jellyfish) → Internal information (Knowledge Hygiene) → Organizational decision-making (Trust and Judgment). Without any one of them, the other two struggle to function independently.

If your answers to all three questions are uncertain—you are not alone. McKinsey's data tells us: fewer than one-third of companies are "mature" in geopolitical risk management [1]; most organizations' knowledge layers are quietly rotting [2]; and the ability to choose between short-term productivity and durable advantage is becoming the next competitive watershed [3][4].

But here is the good news: the moment these three questions are recognized, change has already begun. They are not a death sentence—they are an action list. As Sandra Durth puts it [3]: "Some organizations give people technology tools. Others fundamentally reinvent how work gets done. That difference shows up in performance."

2026 may not be the year AI replaces people. But it is the year companies choose who they are.

Which one is your organization?

Three Questions to Know If Your Company Will Survive 2026