The Escalation Rule Pattern Every AI Agent Needs (and Most Skip)

Source: Dev.to

Why Escalation Rules Matter

When an AI agent encounters an edge case and no escalation rule is defined, it simply guesses. This can lead to serious problems, especially in production where edge cases multiply.

Typical situations that should trigger escalation:

• Low‑confidence writes
• Financial outputs
• First‑time contact

Escalation rules act as a seatbelt: you don’t notice them when everything works, but you’re grateful for them when things go sideways.

Example Escalation Rules

In practice, the rules are a small block in your SOUL.md or AGENTS.md file:

## Escalation Rules
- Escalate to Patrick before sending any external email to a new contact
- Escalate before posting anything that mentions revenue, pricing, or customers
- Escalate if confidence in output is below 0.7 on any write operation

When to Use Them

• Fast development cycles: It’s tempting to skip escalation rules to ship quickly, but production environments demand them.
• Testing vs. production: Scenarios that pass in testing may break in real‑world usage; escalation rules catch those unexpected failures.

Running AI agents in production? The Ask Patrick Library provides battle‑tested configurations for escalation, memory architecture, and cost control.

askpatrick.co