When AI hollows out organizations from the inside

There is a phenomenon Dominican companies are importing without realizing it. Every time they adopt an artificial intelligence tool to reduce operational friction — automating responses, delegating tasks, replacing meetings with agentic workflows — they are also importing a problem that global research has just named with data: organizational loneliness induced by…

There is a phenomenon Dominican companies are importing without realizing it. Every time they adopt an artificial intelligence tool to reduce operational friction — automating responses, delegating tasks, replacing meetings with agentic workflows — they are also importing a problem that global research has just named with data: AI-induced organizational loneliness.

According to a recent study by researchers Constance Noonan Hadley and Sarah L. Wright, published in Harvard Business Review this April, 52% of workers report feeling significantly lonely while working. The figure is not new; what has changed is the emerging cause. AI implementation is acting as a catalyst for that isolation, and it does so through a mechanism the research itself describes with clinical precision: as agentic systems take over routine interactions between colleagues, the exchanges where professional trust is built disappear too.

The mechanism is thermodynamic in its logic. Organizations are systems that sustain their internal temperature — their cohesion — through productive friction: the colleague who asks for help, the team that negotiates a decision, the manager who explains instead of delegating to an agent. When AI removes that friction, the temperature drops. Efficiency goes up, social density goes down. And the balance sheets do not register that cooling until the talent has already left.

The matter goes further. Hadley and Wright document that talking to a system that is always available, never busy, never uncomfortable, progressively atrophies the tolerance for real disagreement. The result, more than apparent loneliness, becomes a social regression. People who work more hours with AI agents develop less capacity to handle the relational complexity that remains exclusively human: creative conflict, operational empathy, loyalty built through shared adversity. One interviewee in their research described AI as «a useful ghost in the office: always there and always responding, but never present.»

That image is more precise than it seems. A useful ghost does no direct harm. It takes up space. It replaces presences. It makes the place look inhabited when in reality it is being emptied out.

For the Dominican context, the warning sign has an additional dimension. Local companies are adopting agentic AI in a labor ecosystem where internal social capital is already fragile — high turnover, generational disconnect, hierarchical structures that concentrate knowledge in a few nodes. On that foundation, replacing human interactions with agents does not optimize the organization: it weakens it. It exports to artificial systems the tactical memory that used to live in conversations between teams. When the agent disappears or the system changes, nothing is left. The company ends up carrying out what in another context I called Corporate Amnesia® by design: not because employees forget, but because the architecture never allowed knowledge to be transferred between people.

In its analysis of the AI ecosystem of 2025 CB Insights identified four priorities for professional services firms competing in the agentic era: orchestrate the agent technology stack, activate proprietary data to make it smarter, turn services into scalable AI products, and lead the building of a human-AI workforce. The fourth priority is the hardest and the least funded. No tool solves it. It requires organizational design, deliberate culture, and the recognition that agents work better when the humans around them work well together.

The concrete question for the Dominican and Latin American business leaders in the middle of generative AI implementations is this: what metrics do you have to know whether your teams' cohesion is rising or falling while operational efficiency rises? If the answer is none, you are optimizing blind. You are measuring the system's speed without measuring the engine's temperature.

The proposal is operationally simple, though culturally difficult. Treat cohesion as a business variable. Design spaces where direct human interaction is structurally mandatory, not voluntary. Audit which conversations existed before an agent was implemented and which disappeared afterward. Create what could be called an organizational heat map: where tacit knowledge flows, which nodes concentrate it, which routes are cut when an agent replaces a conversation.

The organizations that understand that the real competitive advantage remains human — the capacity to build trust, transfer tacit knowledge, adapt with judgment to the unexpected — will be the ones that integrate technology without hollowing out from within. The rest will have gained efficiency and lost substance. A useful ghost in the office, always there, never present.

—

Sources:

  • Hadley, Constance Noonan and Wright, Sarah L. «Employees Are Relying on AI for Personal Support. That’s Risky.» Harvard Business Review, May 2026.
  • CB Insights. «The Future of Professional Services: How firms will capture value in the AI agent era.» CB Insights Research, 2025.