Who Will Absorb the Work?

The future is arriving in a burst of promise, but inside every organization people still have to translate change into practice.

The Human Infrastructure of AI

Everyone’s dad and neighbor can hold at least a brief conversation about artificial intelligence. He might be using ChatGPT to draft emails. He might be investing in AI stocks. He might think the whole thing is overhyped and is ready for it to blow over. He might suspect junior employees are using it to "cheat." There may or may not be conversations swirling about technology ruining school, creative writing, work, art, democracy, or whatever else one cares about.

People may not understand exactly how it works.

But there are opinions.

That alone feels worth noticing. AI has moved out of Hollywood sound stages and into ordinary life. People preface conversations with, “According to ChatGPT…” or “I asked Claude…” before launching into topics like government policies, investment portfolios, or factoids. With varying levels, there appears to be some ambient dread people carry when they know something is changing but cannot yet tell whether it will help them, replace them, expose them, or simply give them one more platform to learn.

But the personal response to AI is not actually what interests me most.

What I keep thinking about is how quickly institutions have started circling it.

Businesses, universities, hospitals, nonprofits, and organizations of every size are trying to figure out how to “leverage AI.” The language is familiar because institutions always seem to find the same words when they are both excited and overwhelmed: streamline, optimize, innovate, transform, scale, improve efficiency, reduce burden, increase productivity.

There is hunger in the language.

I get it. Most organizations are drowning in their own complexity. People are tired. Systems are bloated. Communication is scattered and information lives in too many places. Work moves through too many hands, while everyone is trying to do more with less. We prove, preferably in a dashboard, that what we are doing is measurable, strategic, and aligned.

AI appears to offer a way through. And perhaps, in some places, it will.

I am drawn to gaps in the work. When I see the distance between what a system promises and how it actually functions, my instinct is to reach toward it--to look for the missing connection, imagine a solution, and search for the possibility of a genuine win-win.

I do not experience friction as evidence that something is broken. I experience it as an invitation to become curious.

What is missing here?

What are people trying to accomplish?

Where is a process failing?

What small connection might allow people, work, or the institution to function better together?

That instinct is part of why AI is so compelling. It creates new possibilities for connecting information, reducing repetitive work, recognizing patterns, and helping people move through complexity. It invites us to imagine systems that are more responsive, humane, and perhaps even more joyful to work within.

But it also raises another question.

Who will do the work of turning that possibility into reality?

I suspect many organizations are about to discover that they do not yet know what they do not know.

They know they should not fall behind. They know familiar vendors are embedding AI into platforms they already use. They know consultants are selling roadmaps. They know peer institutions are experimenting. They see the mediafication early adoptors and fast followers.

They know the future is arriving with a flossy new deck. They haven't yet figured out if that deck is developed from their own org or if they're looking at a consultant's.

What they may not yet understand is that adopting AI is not simply a technical project.

It is a cultural one.

Consultants, developers, vendors, and internal technology teams will be brought in to integrate AI into existing operations. They will map workflows, automate tasks, connect systems, build tools, and hold training sessions. On paper, this will look like implementation.

But anyone who has spent time inside a complex organization knows that several versions of work are always happening at once. There is the workflow as imagined by leadership. There is the workflow as documented in policy. There is the workflow as described in the roll out meetings. And then there is the work as it actually happens.

The work as it actually happens is messier. It is more historical. It is more relational. It depends on memory, timing, judgment, personalities, and a thousand small acts of translation that rarely make it into a process map.

It includes the form no one fills out correctly because the instructions are unclear.

It includes the spreadsheet that still exists because the official system never quite worked.

It includes the meeting before the meeting and the parking lot meetings after the "real" meeting.

I'm not easily seduced by the official story. My brain wants to know what people actually do. I pay attention to rituals, hierarchies, symbols, status, language, silence, obligation, memory, and the informal rules people follow while appearing to follow the formal ones.

I'm interested in the gap.

Not merely to criticize it, but to understand it.

Applied anthropology brings that attention into the microcultures of real world of institutions, workplaces, policies, technologies, and human problems. It asks how people make meaning, how they adapt, negotiate change, and how systems are shaped by the people living inside them.

AI adoption is very much a human problem. The public conversation often asks whether AI will replace workers. That question matters. I am not dismissing it. But inside many organizations, I suspect the first and more immediate experience of AI will not be replacement.

It will be absorption.

Someone will absorb the confusion. Someone will absorb the training burden. Someone will absorb the cleanup. Someone will absorb the distance between the vendor’s promise and the institution’s reality. Someone will translate the new system into the old culture.

Often, that someone will not be the person who approved the purchase, announced the initiative, or led the launch meeting.

It will be the high performers: our executive assistants, coordinators, project managers, analysts, program directors, middle managers, and the staff members who already know how everything actually works.

These are the people who know which process is official and which process is real. They know who has to be copied even though the organizational chart says otherwise. They know which deadlines are flexible and which are not. They know who holds the missing information. They know which system people avoid. They know what broke three years ago and was quietly absorbed into somebody’s daily routine.

In many organizations, these people are treated as support. But in practice, they are interpreters. They interpret leadership priorities. They interpret policy, urgency, silence, confusion, and tone. They interpret what a system is supposed to do and what it is actually capable of doing.

They are the human infrastructure.

And when AI arrives, these are the people who will be asked, formally or informally, to make it usable.

They will test the tool. They will clean the data. They will rewrite the instructions. They will re-explain a new workflow. They will notice the exceptions, calm the resistance. They will catch the errors and ask unanswerable questions. They will create the workaround when the tool does not match the work. They may not be called AI strategists. They may not receive a flashy new title. They may not be included early enough to shape the decision. But they will become part of the implementation layer.

This matters because when organizations fail to see that labor, they misunderstand the cost of change.

A new system may save time in one place while creating work somewhere else. A dashboard may make performance more visible while hiding the labor required to produce clean data. An automated communication tool may reduce drafting time while increasing the need for judgment about tone, accuracy, timing, and audience. A chatbot may answer routine questions, but someone still has to decide what counts as routine, what information is safe to automate, when a human needs to intervene, and who is accountable when the answer is wrong.

This is why I keep thinking about AI as a cultural event, not merely a technical tool.

AI will not enter organizations as a neutral object. It will enter workplaces that already have histories, habits, loyalties, fears, bottlenecks, myths, and sacred cows. Every institution has its folklore. Every workplace has the things people say out loud and the things everyone knows but rarely names.

AI will land inside all of that. Then people will begin negotiating with it. They will decide when to trust it and when to double-check it. They will compare its answers to what they already know. They will develop new routines, reject some recommendations, invent workarounds, and gradually determine where the tool belongs within the culture of the organization.

Inevitably, among all this change and excitement teams exposed to new tech will be fearful, enthusiastic, use AI poorly, and yet there will be new discoveries and possibilities no one anticipated.

That process of negotiation is not a failure of implementation. It is implementation.

People do not simply receive technology. They interpret it, reshape it, resist it, accommodate it, and make it meaningful within the world they already inhabit.

Someone will have to help the organization make sense of that process.

That sense-making is work.

It is the labor beneath the workflow: the hidden translation layer between formal design and lived reality. It is the work of turning an idea into a process, a process into a habit, and a habit into something people can actually live with. This labor is difficult to measure because it is not one task. It is the connective tissue between tasks.

It is remembering, following up, or noticing what is missing before the absence becomes a problem. Much of institutional life already depends on this kind of labor, and much of it remains invisible. That is especially true in higher education, healthcare, nonprofits, and government agencies, where official structures are often too slow, layered, or fragmented to account for how work actually gets done, especially to senior leadership.

In those environments, the people who keep things moving are not always the people whose labor gets named. And the better they are at it, the less visible the labor becomes.

That is the trap.

Competence and capacity can make labor disappear.

When someone is good at absorbing ambiguity, they are given more ambiguity. When someone can repair broken systems, they are handed another broken system. When someone can make chaos look manageable, the organization may confuse their effort with ease.

Even the instinct to create a win-win can become its own kind of vulnerability.

The person who sees the gap is often assumed to be the person who should close it. The person who can imagine the connection becomes responsible for making it. The person who believes a better outcome is possible may quietly take on the labor required to produce it.

But solutions do not materialize simply because someone can envision them.

Connections require labor.

And a win-win is only genuinely mutual when the work of creating it is recognized, shared, and supported.

AI could intensify this pattern. Not because AI is inherently bad, but because organizations often adopt tools faster than they understand the labor required to sustain them. A tool can be purchased quickly, but a true workflow takes shape long after the launch.

People have to trust a system, and in order to do so they have to understand it. They have to know when to use it and when not to use it. They have to know what has changed, what has not, and what risks the tool introduces. They have to know whether it solves a problem they actually have or simply creates a new layer of compliance around an old one.

They also need room to be surprised.

One of the most hopeful things about people is that they do not only adapt through compliance. They experiment. They improvise. They discover uses that were not included in the original plan. They connect tools to needs that leadership did not know existed.

This is where change can become more than something imposed.

It can become something shaped together.

But for that to happen, organizations have to approach implementation with curiosity rather than certainty. They have to be willing to learn from the people closest to the friction. They have to see resistance not only as reluctance, but sometimes as information. They have to notice when a workaround is revealing a deeper truth about how the work functions.

If the labor of adaptation is not named, planned for, and resourced, it will fall to the people closest to the friction. But if those people are included early, something else becomes possible. This is where applied anthropology has something to offer.

An anthropological approach to AI adoption would begin with people, not the platform.

It would ask: What work is this tool supposed to reduce? Whose work will become more visible? Whose work may become less visible? What informal practices currently hold the organization together? Where does friction already live? Who usually repairs it? What new labor will the tool create? Who will train, translate, troubleshoot, maintain, and emotionally carry the change?

It would also ask one of the questions institutions often skip when moving too quickly:

What do the people closest to the work already know? The people closest to the work often understand the system best. Not always in the formal language of strategy or technology, but in the lived language of practice. They know where the process bends. They know where the official story fails. They know who the true stakeholder is and how shifts in the work will affect the people we work to serve.

They may also be the first to see what could work. They can identify connections that are invisible from the top. They can recognize where automation would create relief rather than simply move labor around. They can help distinguish between a problem that needs technology and a problem that needs clearer expectations, better communication, or a different decision.

If organizations want AI to work, those people need to be in the room early. Early.

Before an institution declares what AI will fix, it should understand what is actually broken. And before it assumes it knows what the technology will become, it should remain curious about what people might create with it.

That is the human infrastructure of AI.

This may be the only way to take technology seriously. AI may help organizations become more responsive, creative, and efficient. It may reduce repetitive tasks and help people draft, summarize, analyze, organize, and recognize patterns more quickly. It may allow people to spend less time moving information around and more time thinking, connecting, deciding, and creating, ultimately creating measurable and meaningful value. AI will not rescue organizations from the need to understand themselves. That, to me, is the most interesting observation.

The future of AI at work will not be determined only by what the technology can do. It will also be determined by how people negotiate with it, what institutions are willing to learn about themselves, and whether the labor of adaptation is made visible.

It will depend on the translators, repairers, and quiet historians.

If we want to leverage AI, we might ask ourselves to become more curious about the human systems we have built, more honest about the labor holding them together, and more imaginative about what we might create next.