# What stops them from going wrong

*Oversee and adjust*

Your AI agents pause before any sensitive action and wait for your approval. PII is detected and redacted before it reaches the model. Content policies block harmful or off-brand output. Execution limits prevent runaway tasks. Set company-wide policies once and every employee follows them, including future hires.

The guardrails system gives you fine-grained oversight over every action your AI employees take.

Execution limits prevent runaway tasks. Company-wide policies ensure every employee follows the same rules.

The result is an AI workforce that works independently on routine tasks but defers to you on anything sensitive.

## At a Glance

- **0** Sensitive actions without approval
- **Auto** PII detection and redaction
- **Company-wide** Policy enforcement

## Use Cases

### Approve sensitive actions before they execute

Require human sign-off before your employee sends an email, makes a payment, or deletes a record. They pause, notify you, and wait. Nothing irreversible happens without your say.

### Block personal data from leaving your system

PII detection scrubs names, emails, phone numbers, and financial data before they reach external tools or APIs. Stay compliant without manually reviewing every output.

### Enforce company policies automatically

Upload your acceptable use policy or communication guidelines. Your employee follows them on every response without needing reminders or manual review.

### Set execution limits to control costs and risk

Cap how many tool calls, tokens, or API requests an employee can make per task or per day. Prevents runaway automation and keeps spending predictable.

## Features in this Category

- **[Approval Before Action](/en/features/guardrails/approval-gateway)** — Your employee pauses and asks before acting on anything sensitive, so you stay in control without slowing down routine work.
- **[Detect and Redact PII](/en/features/guardrails/pii-detection)** — Personal data like names, emails, phone numbers, and credit cards is caught and redacted before it ever reaches the AI model.
- **[Built-in Content Safety](/en/features/guardrails/content-policies)** — Built-in content filters block harmful or off-brand output, and custom rules let you enforce your own standards.
- **[Execution Limits and Boundaries](/en/features/guardrails/execution-limits)** — Time limits, retry caps, and resource boundaries prevent runaway tasks so you never face surprise costs.
- **[Company-Wide Policies](/en/features/guardrails/company-policies)** — Set policies once at the company level and every employee follows them automatically, no matter how many you hire.