Agents as Employees

The Shift from Assistants to Employees

For years, the dominant framing of AI in the workplace has been “copilot” — a helpful assistant that augments human workers. A new paradigm is emerging: AI agents treated not as tools but as employees. They receive credentials, identity, purchasing authority, and task assignments. They are managed, reviewed, and evaluated the same way human staff are.

This is not a metaphor. Companies are literally assigning agent identities in their systems, giving them access to purchase software, triage customer support, write and deploy code, and manage workflows. The relationship between organization and agent increasingly mirrors the employer-employee relationship — minus the HR overhead.

Key Drivers

1. Autonomous Agent Capabilities

The foundation is technical: agents have crossed the threshold from “needs constant supervision” to “can be trusted with real work.” OpenClaw, NanoClaw, and similar frameworks have demonstrated autonomous coding, customer support triage, and workflow management that operate without human review for routine tasks.

2. Economic Pressure

The math is straightforward. Jerry Murdock (Insight Partners) notes that SMBs see the most immediate impact — a single AI “employee” handling customer service dramatically changes the economics of a 3-4 person company. Klarna’s Sebastian Siemiatkowski demonstrated it at scale, cutting 50% of headcount while expanding into new product lines without additional budget.

3. Consumption-Based Business Models

When agents are the buyers, pricing shifts from seat-based licensing to consumption-based models. Docker, E2B, and others are already moving toward this — the agent uses a service, you pay for what it consumed. This eliminates the human procurement cycle entirely.

Who’s Saying This

Jerry Murdock (Insight Partners):

“Software is going to be purchased or used by agents. An autonomous agent becomes an employee. You give it credentials. You give it identity. And then it’s up to the agent to make the decision.”

Sebastian Siemiatkowski (Klarna):

“I don’t have to deal with sick leave. I don’t have to deal with entitled millennials.” (While provocative, this captures the operational reality of agent-staffed companies.)

Implications

For Software Companies

Your next customer may not be human. Software that cannot be consumed by agents — that requires human-facing UIs, manual configuration, or per-seat licensing — faces obsolescence. APIs, consumption-based pricing, and agent-accessible interfaces become table stakes.

For the Labor Market

Murdock predicts the first impact hits new hires, not current employees. Companies stop hiring junior developers, assistants, and marketing staff. The effect cascades from SMBs (fastest adoption) to enterprise (slowest, but inevitable). He predicts this will be a defining political issue within two and a half years.

For Investors

Evaluating startups now requires assessing how well founders deploy agents. Murdock argues that the quality of a company’s autonomous agent usage will become a deciding factor in venture capital decisions, putting human investors and AI-native startups on the same playing field.

Timeline

Date	Event
2024	Copilot era peaks — agents still assistive
2025	OpenClaw, NanoClaw, and custom agents begin autonomous coding
2026-Q1	Portfolio companies report Cursor “obsolete” — agents write code independently
2026-02	Anthropic’s 16-agent team builds 100k-line C compiler in 2 weeks ($20k)
2026-02	Jensen Huang declares “the market got it wrong” — agents amplify tools, not replace them
2026-H2	Enterprise adoption begins at scale
2028 (projected)	AI labor displacement becomes electoral issue

Related Reading

• AI Agents - The technology enabling the shift
• Agent Orchestration - How agent workforces are coordinated
• Knowledge Work Disruption - The broader labor market impact
• Jerry Murdock - Key voice on agents as employees