Understanding Agent Actions: Add vs. Replace vs. Suggest vs. Do Nothing

When configuring an Automated Agent, you'll choose an action for each field: Add, Replace, Suggest, or Do Nothing.

Think of these as four different levels of automation:

Add = "Fill in the blanks, don't touch what's already there."
Replace = "Update with the latest information, even if something exists, override it."
Suggest = "Show me what you found, I'll decide what to do."
Do Nothing = "Skip this field entirely, don't update it."

Choosing the right action is the difference between a smooth, automated workflow and a frustrating mess of overwritten data.

This guide will help you:

Understand exactly what each action does
Choose the right action for each field type
Avoid common mistakes
Use a simple decision tree to make smart choices

The Four Actions Explained

🟢 Add: "Fill in the Blanks."

What it does:

If the field is empty, the agent adds the new information
If the field already has a value, the agent appends the new information (adds to existing content)

When to use it:

Building content over time (like summaries or notes)
Capturing information that accumulates (like skills or certifications)
First-time data entry for new profiles

Best for field types:

Summary / Long text fields
Multi-select fields (skills, tags)
Notes or comments

🔴 Replace: "Always Use the Latest."

What it does:

Overwrites any existing value with the new information
Works regardless of whether the field is empty or already filled

When to use it:

Information that changes over time (like current role or salary expectations)
Single-point-in-time data (like notice period or availability date)
When the latest information is always the most accurate

Best for field types:

Single-select dropdowns
Date fields
Number/money fields
Yes/No fields
Current status fields

🟡 Suggest: "Let Me Review First."

What it does:

The agent proposes changes, but doesn't apply them automatically
Creates a suggestion that appears on the profile
You review and choose to Accept, Reject, Modify, or Do Nothing with the suggestion

When to use it:

Sensitive or critical information (like salary or contract terms)
Fields where accuracy is paramount
When you're testing the agent's accuracy
When you want oversight before committing changes

Best for field types:

Any field where you want manual review
Financial information
Legal/contractual fields
When you're still learning what the agent captures

⚪ Do Nothing: "Skip This Field"

What it does:

The agent ignores this field completely
No updates, no suggestions, no changes
Field remains exactly as it was

When to use it:

Fields you update manually (don't want automation)
Fields where the agent frequently makes mistakes
Temporary exclusion while testing other fields
When reviewing suggestions: you've seen the suggestion but choose not to apply it

Best for:

Manual-only fields (like internal notes or strategic info)
Fields that require human judgment
When you want to skip a specific suggestion without rejecting it

Quick Comparison Table

Aspect	Add	Replace	Suggest	Do Nothing
If field is empty	Adds new value	Adds new value	Proposes new value	No change
If field has value	Appends (text) or skips (other)	Overwrites existing	Proposes change	No change
Manual review	No	No	Yes	N/A
Best for	Building content over time	Latest info always wins	Sensitive data	Manual control
Your involvement	None	None	Required	Complete control

The Decision Tree: Which Action Should I Choose?

Use this flowchart to decide which action to pick for each field:

Understanding "Do Nothing" in Two Contexts

"Do Nothing" has two meanings depending on where you use it:

1. In Agent Configuration (Setup)

When setting up your agent rules, choosing "Do Nothing" means:

The agent will completely skip this field
It won't analyze the field, won't suggest changes, won't update it
The field is excluded from automation

Example: You configure an agent to update Summary (Add) and Remote Preference (Replace), but you select "Do Nothing" for Salary Expectations because you always enter that manually.

Result: The agent never touches Salary Expectations. Ever.

2. When Reviewing Suggestions

When the agent has made a suggestion and you're reviewing it, choosing "Do Nothing" means:

You've seen the suggestion
You're not accepting it
You're not rejecting it either
You're just... doing nothing with it (for now)

Example: Agent suggests updating Salary Expectations to $120K. You're not sure if that's accurate yet. You select "Do Nothing" and leave it in your Pending Suggestions queue to revisit later.

Result: The suggestion remains pending. Field stays unchanged.

Key Difference:

Configuration "Do Nothing" = Never automate this field
Suggestion "Do Nothing" = Skip this specific suggestion (but agent can suggest again next time)

[Screenshot: Two side-by-side images showing "Do Nothing" in agent setup vs. in suggestion review UI]

Common Mistakes to Avoid

❌ Mistake 1: Using Replace on Summary

What happens: Every call overwrites the entire summary. You lose historical context.

Fix: Use Add for Summary (unless you explicitly want only the latest information).

❌ Mistake 2: Using Add on Single-Select Fields

What happens: If the field already has a value, nothing updates (even if the candidate's answer changed).

Fix: Use Replace for single-select fields.

❌ Mistake 3: Using Replace on Sensitive Data Without Review

What happens: Incorrect salary or contract terms get locked in. Potential legal/financial issues.

Fix: Use Suggest for salary, contracts, and critical dates.

❌ Mistake 4: Using Suggest for Everything Long-Term

What happens: You spend hours reviewing suggestions. Defeats the purpose of automation.

Fix: Start with Suggest, graduate to Add/Replace as you gain confidence. Keep Suggest for judgment fields.

❌ Mistake 5: Not Using Suggest for Strategic Fields

What happens: Agent tries to automate subjective assessments, creates bad data.

Fix: Use Sugges for any field that requires human judgment, complex logic, or strategic thinking.

❌ Mistake 6: Not Testing Your Configuration

What happens: You don't realize the agent is overwriting data until it's too late.

Fix: Run a test call and verify the behavior before rolling out widely.

Pro Tips

💡 Tip 1: Use Different Actions for the Same Field in Different Rules

Example:

Rule 1: Summary → Add (for candidates in "Screening" stage)
Rule 2: Summary → Replace (for candidates in "Archived" stage who you're reactivating)
Rule 3: Summary → Suggest (for VIP candidates where you write custom summaries)

All three rules can coexist. AIRA applies the one that matches the criteria.