When Words Fail

When Words Fail

Why the old organizational metaphors cannot describe what AI is doing

by Mike Herzog, President, the telos institute

This essay is the first in a series exploring how AI is destabilizing the ways organizations understand themselves, and what kind of leadership may be required when the old frames no longer hold.

The Frames We Live By

For hundreds of years, organizations have made sense of themselves through metaphor.

Management theorists and organizational scholars have named dozens of them: organizations as families, armies, cultures, brains, political systems, psychic prisons, ecosystems, networks, and more. Gareth Morgan’s Images of Organization remains one of the clearest demonstrations of how many such frames are available. This essay is not a comprehensive taxonomy. I am focusing on three broad frames that have shaped modern management thinking: machine, organism, and living system.

During the Industrial Revolution, when organizations needed to coordinate thousands of workers across multiple shifts, they came to understand themselves as machines. Logical and linear. The metaphor answered the most pressing question of the time: how do we optimize efficiency and productivity?

The mid-twentieth century brought rapidly changing markets, accelerating technology, global competition, and shifting social expectations. These pressures fractured the rigid machine model. A new metaphor emerged, one that favored adaptation over precision. Organizations became organisms, centered around a unifying imperative: survival. Organisms could thrive, but they could also fall sick. The same patterns that kept the system alive could also keep it diseased.

The late twentieth century was marked by the rise of complexity. Outcomes could no longer be reliably planned, predicted, or controlled from the top. The boundaries of the organism metaphor blurred as organizations found themselves entangled with suppliers, global competitors, customers, regulators, capital markets, and ecological constraints. Seeing organizations as living systems accommodated this complexity: less designing the perfect structure, more cultivating the conditions that allow healthy patterns to emerge.

These metaphors are not merely labels applied retrospectively. They are human responses to real environmental pressure. Once named, they become more than descriptions. They become interpretive lenses, diagnostic tools, and prescriptive models that organizations actively organize toward. This is the deeper point Lakoff and Johnson made in Metaphors We Live By: metaphors are not merely decoration, they are definition.

A romanticized view of history tells us these stages evolved in a clean, linear progression: the limits of one metaphor get exposed, a new one emerges, and the world gets better. Reality is messier. Most organizations never made the leap. The machine metaphor still dominates the global landscape. Organism thinking has taken hold in some quarters. Living-systems thinking, despite thirty years of careful work by Senge, Wheatley, Scharmer, and others, has remained mostly the province of the leading edge.

The metaphors an organization lives by shape everything: strategy, product, culture, and posture toward social and environmental questions. They reveal where the organization is likely to thrive and where it is likely to struggle. Most importantly, they govern how the organization moves into the unknowable future.

What AI Does to Them

The emergence and acceleration of generative AI is calling into question every layer of our work, from the individual to the enterprise to the global marketplace. It does not merely automate tasks. It challenges what has long felt like a human monopoly: the ability to interpret, decide, generate, and increasingly act.

We can debate the pace of arrival, but the direction is no longer in doubt. AI will bring change at a scale we have not encountered before. It sits in a liminal position between tool and actor, and we do not yet have adequate language for what that means.

But organizations do not all make sense of the world through the same frame. The machine, organism, and living-systems metaphors still coexist, often inside the same company. Under pressure, one of them usually becomes easier to see.

So how will these different frames shape the way organizations and their leaders make sense of AI?

For organizations running on the machine metaphor, AI looks like jet fuel in the tank. Personal efficiency, process automation, and deeper analytical insight are exactly what these organizations have always wanted. AI makes all of it easier. From inside this frame, the path appears obvious: manage the risks, train the workforce, automate what can be automated, and move full steam ahead.

The machine frame has always had one deep blind spot: intimacy. I do not mean warmth, friendliness, or interpersonal closeness. I mean the deep understanding that comes from relationship, context, trust, history, embodied judgment, and mutual recognition. Machines can process, sequence, optimize, and execute. But machine-framed organizations tend to make processes matter more than people.

AI will not repair this weakness. It will intensify it. In machine-framed organizations, AI will be pulled toward efficiency: fewer handoffs, fewer meetings, fewer people, faster decisions, cleaner workflows. Some of that will be useful. But the same logic will also strip away the human interactions through which trust, judgment, apprenticeship, and shared understanding are built. The relational fabric of the organization, already thin in machine-framed environments, gets thinner.

The organism frame reads the same technology through a different question. Not "how will this help us?" but "how might this kill us?" The existential threat to organisms is the erosion of boundaries. AI dissolves them everywhere at once: between functions, between roles, between human judgment and machine output, between the organization and the market it once stood apart from.

To save themselves, some organizations will outright reject the technology. They will retreat into protective boundaries until the technology feels safe. But most organism-framed organizations will not openly reject AI. The pressure to adopt is too strong, so they will do something more subtle. They will adopt AI on the surface while resisting it at a more cellular level.

This resistance is not always cynical or conscious. Often it is simply how organizations protect coherence when change exceeds their capacity to metabolize it.

Pilot programs that never scale. Centers of excellence that produce slide decks but not change. Tools deployed in narrow workflow corners where they cannot disturb the organizational structure. Adaptation on the surface, containment underneath. The containment cannot work, but the trying buys time.

Living systems thinking emerged in response to extreme complexity, in the work of Wheatley, Senge, Scharmer, and others. It was built for environments the machine and organism frames could not read.

For that reason, it should seem the most capable of accommodating AI. AI is complex, adaptive, distributed, and nonlinear. It moves through networks rather than chains of command. It crosses boundaries rather than preserving them. It produces patterns no single actor fully controls. If any inherited metaphor should be able to make sense of this technology, it would be the living-systems frame.

But AI exposes a deeper vulnerability in the living-systems frame. AI is not alive, but it is no longer inert in the way our tools used to be. It learns, responds, adapts, converses, remembers, anticipates, and participates in the organization’s patterns of perception and action. It changes what people notice. It alters what counts as knowledge. It reshapes feedback loops, authority, coordination, and memory. AI becomes a nonliving participant inside a living system: active, but not alive.

This is what makes the living-systems frame becomes expossed. Its promise rests on wholeness, emergence, relationship, and distributed intelligence. AI can simulate all four. It can make the organization appear more connected, more intelligent, and more coherent than it really is.

The danger is not simply that AI accelerates the system. It is that AI optimizes parts of the system in ways that can quietly degrade the judgment, friction, ambiguity, and relationship on which the whole system depends.

Knowledge moves faster than ever before, available to anyone who asks, while the human relationships that once carried that knowledge begin to thin. Learning appears to accelerate, while individual judgment quietly atrophies into dependence on outputs no one quite understands. Each local team optimizes its work brilliantly, while the long-term adaptability of the whole system erodes in ways no one is tracking. Coordination across teams looks more intelligent, even as the power structures shaping that coordination become harder to see, harder to name, and harder to challenge.

A company uses AI to synthesize every meeting, customer conversation, Slack thread, support ticket, and strategy document. On the surface, the organization now has more shared intelligence than ever. Leaders can ask the system anything. Teams can see patterns across the whole enterprise. Knowledge becomes available instantly.

From a living-systems perspective, this can look like greater wholeness.

But what may actually be happening?

People stop carrying knowledge relationally. They stop talking across boundaries because the system “knows.” Judgment moves from embodied experience into generated summaries. The organization appears more connected, while the human relationships that make real adaptation possible quietly weaken. The AI has produced a representation of the whole, not the lived coherence of the whole. This is counterfeit wholeness.

The living-systems frame is therefore not simply more ready for AI than the machine or organism frames. It is more exposed to AI’s deepest ambiguity. AI is not alive, but it mimics aliveness by learning, conversing, adapting, patterning, and participating. The frame sees the right terrain, but AI gives it a counterfeit version of what it has been seeking.

This is not only a conceptual problem. It is a leadership problem. 

If AI is moving organizations into liminal space, the first leadership task is not to rush toward a new metaphor. It is to see which old metaphor is still governing the organization’s response.

Reading the Organization Through AI

Leaders do not consciously choose the metaphor their organization lives by. The metaphor is usually hidden beneath awareness. It lives in assumptions, habits, budgets, org charts, meeting patterns, incentive systems, and the private reflexes of decision-makers. Leaders may not say, “we are a machine,” “we are an organism,” or “we are a living system.” But the organization will still behave as if one of those frames is true.

AI makes these hidden frames easier to see by using this diagnostic question: What does AI cause us to protect, accelerate, automate, resist, or pretend not to see?

If the organization treats AI as a way to increase output, reduce friction, automate coordination, and extract more productivity from the same system, the machine frame is probably operating. The risk is not efficiency itself. The risk is that every form of human friction starts to look like waste. Trust, judgment, care, apprenticeship, conflict, and conversation become costs to be minimized rather than capacities to be protected. Intimacy dies at the hands of efficiency.

If AI is met with repeated appeals to quality, standards, risk, brand, compliance, role clarity, or “the way we do things here,” the organism frame may be operating. This will not always look like fear. It may look like stewardship. The organization will say it is protecting what matters, and in many cases it will be. But leaders should watch for the moment when necessary protection hardens into an immune response. The signs are subtle: pilots that never evolve, governance processes that expand faster than learning does, approved use cases that avoid the organization’s real work, and language about safety that quietly becomes language for preserving the existing order. The organization has not rejected AI. It has metabolized AI in ways that prevent AI from changing it.

If AI provokes both fascination and distrust, the living-systems frame may be operating. This response is less straightforward. Some leaders will see AI as a way to reveal patterns, connect disparate parts of the enterprise, and accelerate learning across the whole. Others will see it as a threat to the very capacities that make living systems alive: embodied judgment, local context, relational knowing, dialogue, trust, and meaning-making. Both responses may be legitimate. The risk is not simply adoption or rejection. The risk is failing to distinguish between real coherence and the appearance of coherence. A representation of the whole is not the same as the lived coherence of the whole. People carrying knowledge relationally is not the same as system knowledge. Embodied experience is not the same as generated summaries. Better coordination will not make up for weak relationships.

The point is not to classify the organization. It is to read it. AI will disclose the frame by revealing what the organization reaches for first: speed, protection, or coherence. Each reflex will illuminate part of AI’s promise. Each will also create its own distortion. Leaders who can see both will be less likely to confuse productivity with progress, stewardship with containment, or connectedness with wholeness.

The Threshold

Each inherited metaphor gives organizations a different way to make sense of AI, and each produces a different failure mode.

Under ordinary conditions, this is how organizational evolution works. If the failures associated with the current metaphor become too costly, the organization may begin to evolve toward a different metaphor.

But there is a fundamental difference between what we are experiencing with AI and the metaphor shifts that came before. AI's pace, scale, and the fundamental fact that it is not “alive” are fracturing all metaphorical frames at once. AI does not just strain each historical frame. It attacks the defining essence of each frame. AI exposes the cold, mechanical nature of the machine-frame. It exploits the boundary-identity that is essential to the organism-frame. It challenges the very essence of the living-systems frame, aliveness itself.

Does this mean we are about to witness the emergence of a fourth metaphor? Perhaps. It is possible we will devise a metaphor that transcends and includes the previous three, preserving their strengths while moving past their limitations.

AI may be pushing us into territory where metaphor itself begins to fail. Humans will keep reaching for metaphor; we cannot do otherwise. But AI moves across the very distinctions metaphors depend on: human and machine, organic and synthetic, alive and not-alive. Metaphors require distinction. Distinction is what is being lost. 

When a metaphor fails, an organization does not simply lose a way of speaking. It loses a way of seeing, deciding, and moving. The old frame no longer explains what is happening, but no new frame has yet become trustworthy enough to guide action. Leaders are left in a condition more difficult than mere uncertainty. The uncertainty is not only about what comes next. It is about what is happening now.

This condition has a name. It is liminal space: the territory between what is dissolving and what has not yet formed.

Liminal space is not new. Anthropologists studied it in the rituals of pre-modern cultures, in the moments between one identity and the next. Developmental theorists have studied it as the territory between stages, when the old structure of self has loosened and the new one has not yet stabilized. Contemplative traditions have studied it for centuries, as the place where ordinary knowing falls away and something else becomes possible.

What is new is the scale on which we are now being asked to inhabit it. Liminal space has historically been the territory of individuals during transitions, or of small groups during rites of passage. Whole organizations, whole industries, whole societies are now being moved into it at once. The dissolution of the metaphors we have used to make sense of organizations is not happening in pockets. It is happening across the field.

This is the threshold we are crossing. The leaders most able to cross it will be those who can see not only what AI makes possible, but what their inherited frames prevent them from seeing.