Skip to main content
Guardrails are the rules that define the boundaries of your AI employee’s behavior. They tell your AI employee exactly what it is allowed to do, what it must hand off to a human, and how it should respond when something falls outside its scope. You define these rules — not Jobr. Your AI employee will never act outside the limits you approve, which means you stay in control of your customer relationships, your commitments, and your business policies at all times.

​
What guardrails control

Guardrails cover three main areas of your AI employee’s behavior:
  • Off-limits topics and actions — Certain subjects or requests are flagged as out of scope. When your AI employee encounters one of these, it does not guess or improvise. It follows a defined response, typically acknowledging the question and routing it to the right person.
  • What the AI can commit to vs. what requires a human — Some actions, like quoting a published price or confirming a booking, are safe for your AI employee to handle independently. Others — like agreeing to a custom deal or issuing a large refund — require a human decision. Guardrails draw that line clearly.
  • Out-of-scope responses — When something falls outside the guardrails, your AI employee responds in a way you’ve pre-approved: politely, professionally, and without leaving your customer hanging. It might say it needs to check with someone, or it might immediately escalate to your team.

​
Examples

Your AI employee can quote the prices listed on your website or in your approved price sheet. It cannot agree to custom discounts, match a competitor’s price, or make any pricing promise that falls outside your published rates. If a customer pushes for a deal, the AI acknowledges the request and routes it to a human who has the authority to decide.
Your AI employee can process refund requests up to a dollar threshold you set — for example, any refund under $200 can be approved and processed automatically. Requests above that threshold are paused and escalated to a member of your team, along with a full summary of the conversation so the handoff is seamless.
Certain categories of questions are always routed to a human, regardless of how the question is phrased. Common examples include legal disputes, regulatory enquiries, formal complaints, and any situation where a customer has indicated they are distressed or considering legal action. Your AI employee will never attempt to handle these — it will acknowledge the customer and pass the conversation on immediately.

​
How guardrails are set

Guardrails are defined during the role mapping phase at the start of your engagement. Jobr works with you to understand your business, your risk tolerance, and the decisions you’re comfortable delegating to an AI. Together, you document the rules, the thresholds, and the response behaviors — and those become the operating boundaries for your AI employee before it handles a single task. You are not locked into those rules forever. If your policies change, or if you want to expand or tighten what your AI employee can do, you can request an update at any time through your Jobr account manager. Changes are reviewed, documented, and applied to your AI employee’s configuration before they take effect.
Your AI employee will never act outside the guardrails you’ve approved. When in doubt, it escalates rather than guesses — every time.

Escalation Rules

Learn how to define when and how your AI employee hands off to a human.