Guardrails are the rules every variant the Decision Agent generates is written within. Set them once per account, and each variant arrives in Content Review already in your voice and clear of your no-go topics.
You set guardrails before your first Content Set. If your account already has a Business DNA, Insider One fills in the matching values, and you can customize them for the Decision Agent without changing your Business DNA. For the bigger picture, see Content Sets in Decision Agent.
In this article:
Set up guardrails
In the InOne panel, navigate to Agent One > Agent One Teams > Content Set > Guardrails.
Review the values in the From Business DNA block. These values are shared across Insider One and appear greyed out.
.jpg)
To change them for the Decision Agentonly, switch on the Customize toggle. Your Business DNA stays as it is.
To edit the Business DNA values themselves, click Edit Branding Settings at the top of the page.
Fill in the Business Context fields.
Fill in the Brand Voice fields.
Fill in the Content Rules and Restrictions fields.
Guardrail fields
Guardrails come in three groups. Fields marked (Business DNA) come from your account's Business DNA.
Business Context
Business Context tells the Decision Agent what it needs to know about your business to write relevant copy.
Field | What to put in | Example |
|---|---|---|
Vertical | Your industry, picked from the list | Choose from the list |
Primary Goal | What a good outcome is for your messaging | Repeat purchase and subscription retention |
Seasonal Context | The moments that change your messaging | Spring SPF, gifting in November and December, routine reset in January |
Positioning (Business DNA) | How you describe the brand in one line | Value-oriented casual apparel with a socially conscious tagline |
Value Proposition (Business DNA) | Why customers choose you | Filled in from Business DNA |
Brand Pillars (Business DNA) | The few themes every message can draw on | Affordable, casual apparel, everyday wear |
Brand Voice
Brand Voice describes how you sound. This group most often decides whether a variant gets approved.
Field | What to put in | Example |
|---|---|---|
Emoji Usage | How much emoji the copy can use | Choose from the list |
Voice and Tone | The one or two tones that are yours | Friendly |
Call to Action Style | How your buttons and links read | Choose from the list |
Urgency Level | How much time pressure is acceptable | Low |
Suggested Keywords | Phrases you like to see in your messages | Add to basket, Continue reading |
Be concrete. "Friendly, plain-spoken, not pushy" in Voice and Tone, paired with the words you never use in Banned Phrases, produces better variants than "friendly tone."
Content Rules and Restrictions
Content Rules and Restrictions are the rules every generated variant must follow.
Field | What to put in | Example |
|---|---|---|
Banned Phrases | Words and framings you never use | Luxury, premium, exclusive, hustle |
Forbidden Topics | Subjects the copy must never touch | Medical or cure claims, competitor mentions |
What Insider One handles for you
Some rules don't need a guardrail field. Here is what Insider One applies automatically.
Group | How it applies |
|---|---|
Delivery constraints | Your account's frequency cap and each channel's silent hours and send windows apply as configured. Send Time Optimization overrides the frequency cap the same way it does on a regular campaign. You don't set them here. |
Offers and codes | The Decision Agent keeps whatever offer is in the base content exactly as written. Variants never invent a code, a gift, or a discount, and never change a value. To include a code in the variants, put it in the base content. |
How guardrails are enforced
The Decision Agent checks every generated variant against your Brand Voice and Content Rules. Each variant receives an AI Match Level of High, Medium, or Low, which you see in Content Review.
A Low match doesn't block a variant. It tells you to look closely. You stay in control: you approve or reject each variant yourself.
Good practice
Fill in guardrails before your first Content Set. A set generated on empty guardrails is a set you will mostly reject.
Put each rule where it is enforced. A wording ban belongs in Banned Phrases, not in a review comment.
Keep journey-level rules out of the account guardrails. "Abandoned cart never carries an incentive" is a label choice for that Content Set, not an account rule.
Review your guardrails after your first review round. The rejections you keep making are the rules you haven't written yet.
Next steps
With your guardrails in place, the Decision Agent has the brief it needs to write variants your team can approve with confidence. Ready to see them in action? Create your first Content Set, then check the AI Match Level on each variant in Content Review.