Radical Reads

In this guest post, George Sivulka, Founder and CEO of Radical portfolio company, Hebbia, argues that in the same way the Internet redistributed the geography of where work gets done, AI is redefining who performs tasks. He explores the rise of “agent employees” — autonomous systems set to fundamentally reshape how organizations staff and operate.

The internet decoupled labor from geography. Now, AI is decoupling labor from humans. 

Just as the internet enabled remote work and catalyzed a shift in organizational design, AI will give rise to “agent employees” — autonomous systems capable of executing tasks once reserved for humans, or previously unimaginable altogether. 

We’re at the precipice of a new divergence in how organizations structure their labor, and the effects are already starting to take place. 

From Place to Personhood

The internet, as a general-purpose technology, transformed the cost and coordination structure of organizations. Remote collaboration tools, cloud platforms, and asynchronous communication allowed firms to separate the performance of work from physical co-location. This enabled three common organizational archetypes: fully in-office, hybrid, and fully remote.

Each firm adopted its model by weighing trade-offs relating to the benefits of working together, managerial oversight, and cultural cohesion.

AI is driving a similar fracture — not in where labor takes place, but in who or what performs it.

Enter the Agent Employee

“Agent employees” are AI systems embedded into business workflows. These are not just tools; they are operational primitives, autonomous or semi-autonomous actors within an organization. 

They come in various forms:

• Reactive agents (e.g., RPA bots)
• Deliberative agents (e.g., planning LLMs)
• Cognitive agents (e.g., agents with memory and self-reflection)
• Multi-agent collectives (e.g., coordinated agent swarms)

These agents can perform knowledge work, synthesize information, draft outputs, and execute decisions — at near-zero marginal cost and unbounded scale.

The Organizational Trichotomy

Thus, we are entering an era where organizations will split along a new axis:

1. Fully Human Organizations
   • Creative industries, bespoke services, high-trust domains.
   • Agents may be used for tooling, but not labor substitution.
2. Human-Agent Hybrid Organizations
   • Centaur models and HITL systems.
   • AI augments human cognition, enabling new forms of productivity.
3. Fully Agentic Organizations
   • AI-native firms where labor is largely automated.
   • Human oversight is strategic, not operational.

These structures mirror the internet-era divergence around remote work, but AI’s paradigm shift is deeper: it redefines cognition, not just coordination.

Implications: Designing for Post-Human Cognition

This transformation isn’t simply about automation — it’s about where cognition resides, how accountability is structured, and what it means to trust non-human contributors. Just as some firms thrived in the shift to remote-native models while others recommitted to the office, so too will organizations differentiate based on how — and if — they integrate agentic labor.

Agent employees challenge the foundations of organizational design:

• Where is judgment exercised?
• Who — or what — drives execution?
• How do we govern intelligence that doesn’t think like us, but works for us?

The internet redefined where work happens. Agent employees are redefining who gets to do it. Some firms will remain fully human. Some will go hybrid. Others will scale through networks of intelligent, autonomous agents.

Across every path, we’re not just redesigning workflows — we’re reimagining what makes a team.

The age of organizational divergence has begun — again.