Custom Agents
How It Works

Autonomy Modes

Draft mode, auto-respond, escalation rules — how Custom Agents lets you build trust gradually, just like you would with a new hire.

How It Works

Configurable autonomy

Businesses do not hand full autonomy to a new hire on day one. They should not hand it to an AI agent on day one either. Custom Agents provides a graduated path from full human review to full autonomy.

Three modes

Draft Mode

The agent writes every response. You review and approve before it sends. Like a new VA's first week — you check every outgoing message.

Auto-Respond

The agent handles routine inquiries autonomously. You get notified only for escalations. Like a trusted VA handling day-to-day operations.

Escalation Rules

The agent operates autonomously but defers to you under specific conditions you define. The guardrails for autonomous operation.

The graduated path

The natural progression mirrors how every business works with a new hire:

WeekModeWhat happens
Week 1Draft ModeAgent drafts every response. You review 100% of outgoing messages. You learn the agent's style and correct mistakes.
Week 2-3Auto-Respond + Strict RulesAgent handles routine emails autonomously. Escalates anything involving money, complaints, or VIPs. You review ~20% of messages.
Month 2+Auto-Respond + Relaxed RulesAgent handles most communication independently. Escalates only genuine edge cases. You review ~5% of messages.

Escalation rules

You define the conditions. The agent follows them exactly:

  • Dollar thresholds — "Escalate any refund request over $200"
  • Topic triggers — "Escalate anything mentioning 'legal', 'attorney', or 'lawsuit'"
  • VIP contacts — "Always escalate emails from @bigclient.com"
  • Sentiment — "Escalate if the customer seems angry or frustrated"
  • Uncertainty — "Escalate if you're not confident in your response"

When an escalation triggers, the agent:

  1. Notifies you immediately (email, text, or dashboard notification)
  2. Provides a summary of the situation and recommended response
  3. Waits for your guidance before responding to the customer

Why this matters

Fully autonomous AI is powerful but dangerous. The ability to gradually increase autonomy as trust builds — starting with draft mode, moving to auto-respond with escalation rules, eventually reaching full autonomy for routine interactions — is what makes Custom Agents viable for real business adoption.

This is not optional safety theater. It is a core product differentiator. Businesses will not trust a fully autonomous agent with zero oversight. Custom Agents gives them the controls they need to feel confident.

Start with draft mode, scale to full autonomy

You control how much independence your agent has. Adjust anytime.