Glossary
What are AI guardrails?
AI guardrails are the rules, limits and checks that keep an AI system within what you allow: topics it must not discuss, claims it must not make, actions it may not take, and points where it must stop and hand over to a person. For a sales agent they decide whether it is safe to let it write in your name.
Published by Sellinger AI. Numbers and settings checked on 24 October 2026 against the fact register. Report an error.
What kinds of guardrails does a sales agent need?
| Type | Example |
|---|---|
| Content rules | Never state a price that is not in the knowledge base |
| Capability limits | Write down exactly what the product does and does not do |
| Hand over rules | Stop and notify a person on any complaint, legal question or discount request |
| Sending limits | Daily caps per account and quiet hours |
| Suppression | Never contact someone who opted out |
| Approval | Every message waits for a person, at least at first |
Why are rules not enough on their own?
Because no list of rules covers every situation. A buyer can ask something nobody expected, and a model can read a rule more loosely than you meant. Guardrails reduce risk; reading real conversations and keeping a person in the loop catches what they miss.
How do you write good guardrails?
- Write them from real conversations, not from imagination.
- Be specific: name the topic, the claim or the trigger.
- Say what to do instead: hand over, ask a question, or decline politely.
- Review them when you see a reply that went wrong.
How Sellinger handles it
A campaign can run in approval mode, where every message waits for a person to approve it, or send automatically. A person can take over any conversation, and the agent stops on its own when it cannot answer, for example a price it has no answer for. Every connected account has daily limits that the agent cannot exceed, set per account. LinkedIn defaults are 20 invitations a day and 150 a week. Email defaults to 50 sends a day per mailbox, with an optional warm up that starts at 5 a day and doubles every 3 days up to the limit.
Before every send on email, LinkedIn and WhatsApp, the platform checks the team's suppression list and refuses the send when the contact is on it or when the check cannot be completed. The write rules your agent follows help article shows the rule editor, and the capability guardrails playbook a real example.
Common questions
Are guardrails the same as approval?
What is the most important guardrail?
Can guardrails make a message legally compliant?
Sources and method
General AI practice. Sellinger behaviour from the fact register, checked 24 October 2026.
Related: AI hallucination, Human approval workflows, Write rules your agent follows, Outreach rules.
