How AI One-to-One Email works
AI One-to-One Email lets the agent generate personalized one-to-one email content at runtime. You configure core delivery controls — sender identity and failure behavior — in the builder, while the agent tailors the subject, preview text, and body to the current situation. This tool combines the personal feel of one-to-one messaging with the flexibility of AI-generated content. It is ideal for high-touch outreach scenarios where the message should reference specific user context — like recent activity, account health, or lifecycle stage — while still appearing as a direct email from a real team member. The agent adapts every send to the recipient, so no two emails need to read the same way.Configuration
Inputs
Instructions that tell the agent when this tool should be used and what quality bar the generated copy should meet.
Target user to email.
Team member identity used as sender.
AI-generated subject line.
Optional AI-generated preview text.
Optional CC recipients; can be set by builder defaults or agent at runtime.
Optional BCC recipients; can be set by builder defaults or agent at runtime.
Controls whether the run should continue when send fails.
Outputs
The status of the email send operation. Can be
success, failure, or skipped.The ID of the user that the email was sent to.
The team member ID used as the sender for this email.
The ID of the email that was sent (available on success).
The options used when creating the email, including the AI-generated subject, preview text, and body content (available on success).
The response from the email creation service (available on success).
The code indicating why the email send operation failed or was skipped. Can be
resend_error, duplicate_email, or user_marketing_email_opt_out.The message of the error that occurred (available on failure).
Use cases
Trial-to-Paid Conversion Follow-UpBest practices
Prompt Quality and Tone Control- Give concrete voice, policy, and CTA constraints in the tool prompt to keep generated outreach consistent and safe.
- Include examples of ideal tone and length so the agent has a clear quality bar for every send.
- Reference specific user properties the agent should weave into the message (e.g., company name, recent milestones, plan tier).
- Configure sender selection rules to ensure the agent uses the correct team member identity for each recipient.
- Avoid generic sender assignments — the
from_idshould map to the person who actually owns the customer relationship. - Audit
from_idandcreate_email_optionsoutputs periodically to verify the agent is sending from the right person with appropriate content.
- Start with a small cohort to evaluate generated copy quality before rolling out to a broader audience.
- Set
complete_on_send_failureappropriately — enable it when the agent run has other critical steps that should proceed regardless. - Monitor
codeoutput for patterns likeduplicate_emailto detect and prevent repeat outreach to the same user.