INTEL-15: The Delegation Threshold

The INTEL  

77% of enterprise AI interactions are now full task delegation — not collaboration. The human is leaving the loop.

Anthropic’s Economic Index data reveals a critical inflection point. Over eight months, directive conversations — where users delegate complete tasks to AI rather than collaborating on them — jumped from 27% to 39%. For the first time, automation usage exceeds augmentation in enterprise settings. Business API usage shows an even starker pattern: 77% of enterprise interactions are automation patterns.

This has direct security consequences. When organizations augment work with AI, humans remain in the decision loop. They review outputs, validate choices, and catch errors — including security-relevant ones. When organizations automate with AI, the agent makes decisions and executes actions that may not receive meaningful human review before reaching production.

Anthropic’s Agentic Coding research describes a “collaboration paradox” — engineers use AI in roughly 60% of their work but can only fully delegate 0-20% of tasks. That gap is the current safety margin. As agent capabilities improve and organizations grow more comfortable with delegation, that margin compresses. The security implications scale with the delegation.

When 42% of organizations trust agents to lead development with limited human review, the boundary between “well-defined task” and “security-relevant decision” is drawn by the agent, not the human. The agent decides which dependencies to use, how to handle authentication, what data to access. These are security decisions being made at machine scale without security oversight.

Why It Matters  

The shift from augmentation to automation changes who is accountable for security-relevant decisions. In a collaboration model, a human reviews agent output and owns the outcome. In an automation model, the agent acts and the human may never see the result until something breaks — or is breached.

This risk compounds with scale. A single developer collaborating with AI makes perhaps dozens of security-relevant decisions per day and reviews most of them. An automated multi-agent architecture handling CI/CD pipelines, customer workflows, and cross-functional processes makes thousands of such decisions — with review coverage that decreases as delegation increases.

The 0-20% full delegation rate represents today’s safety margin. It will not hold as agents improve. Security programs need to be architected for a world where delegation is the default, not the exception.

What To Do  

Treat the automation threshold as a security architecture priority — build review gates where agent autonomy meets production impact. This doesn’t mean reviewing every agent action. It means defining the blast radius of each agent’s autonomous scope and requiring human authorization for actions that exceed it. Classify agent tasks by security impact: low-impact actions (formatting, documentation, boilerplate) can run autonomously; high-impact actions (deployment, credential access, data pipeline changes) require human checkpoints. The Agentic Coding report’s concept of “intelligent collaboration” — agents that learn when to ask for help — should be a security design requirement, not just a productivity feature.

MITRE ATT&CK  

• T1059 — Command and Scripting Interpreter: AI coding agents with autonomous execution are, by design, command interpreters — legitimate execution paths that can be exploited if agent behavior is compromised
• T1053 — Scheduled Task/Job: Long-running autonomous agents in CI/CD pipelines create persistent execution mechanisms that blend legitimate automation with potential compromise vectors

Learn More  

• Anthropic 2026 Agentic Coding Trends Report — Primary source
• Anthropic 2026 State of AI Agents Report — Primary source (available from Anthropic)
• FIR Risk INTEL-14 — Shadow Agents — Machine identity governance gap
• FIR Risk Tuesday E87 — The Agents Have Keys — Full dual-report analysis

Powered by FIR Risk Platform — AI-driven threat intelligence for enterprise risk leaders.