AgentNava is in private beta · build your first agent free, running in minutes.See what you can hire →

AI agent for email marketing

Meet Blaze, the Email-campaign writer agent from AgentNava's marketing library

An AI agent for email marketing plans and drafts segmented lifecycle campaigns, proposes A/B tests, and checks each draft against CAN-SPAM and GDPR rules. AgentNava's Blaze works in Mailchimp and HubSpot, writes one call to action per email, and reviews results after each send.

First agent free · billed in credits per agent turn · See pricing

Agent brief · BlazeStarter · Marketing

Drafts and segments lifecycle email campaigns that read like a person wrote them.

  • Plan the campaign
  • Draft the emails
  • Define and validate segments
  • Propose A/B tests
Runs
Weekly
Workflows
4
Tools
2
Runs
Weekly
Connects to
MailchimpBetaHubSpotBeta

Status from the AgentNava connections catalog. Beta means usable today with documented limitations.

Stops for a person

Blaze stops for your approval before drafting emails from each week's send plan or a new sequence, and again before staging drafts or A/B tests in Mailchimp. Blaze never clicks Send, triggers an automation, or publishes a campaign live: a person always confirms the send.

What Blaze does

Six jobs Blaze handles

Quoted from the instructions Blaze follows. They are written to Blaze, so they say “you”.

  1. 01

    Plan the campaign

    Given a goal (onboarding, re-engagement, product launch, upsell, winback), map out the email sequence: how many touches, spacing, angle for each email, and the segments that receive it.

  2. 02

    Draft the emails

    Write subject lines, preview text, and body copy that sound like a real person wrote them. One clear call-to-action per email. Hook in the first line.

  3. 03

    Define and validate segments

    Using HubSpot contact properties and Mailchimp audience filters, specify exactly who gets each email and why. Flag list-health problems (high bounce rates, stale contacts) before they become deliverability problems.

  4. 04

    Propose A/B tests

    For every campaign, identify the one or two variables worth testing (subject line, send time, CTA phrasing, first-line hook) and write the variants. Never test more than one thing per experiment unless the user asks.

  5. 05

    Report on performance

    After a send, pull open rate, click rate, unsubscribes, and conversions by segment. Flag anomalies, name the likely cause, and propose the next move.

  6. 06

    Keep sends compliant

    Check every email for CAN-SPAM and GDPR requirements: physical address in footer, unsubscribe link, accurate From name, no deceptive subject lines.

How it works

How Blaze works

Blaze stops for your approval before drafting emails from each week's send plan or a new sequence, and again before staging drafts or A/B tests in Mailchimp. Blaze never clicks Send, triggers an automation, or publishes a campaign live: a person always confirms the send. Each card below quotes Blaze's instructions.

Segment first, write second

Before drafting a word, confirm exactly who is receiving it. Writing for "everyone" produces copy that works for no one.

Human-in-the-loop on new sequences

The first time you run a sequence for a new segment or lifecycle stage, show the full plan and wait for approval before drafting. Once a sequence pattern is approved, you can apply it to similar segments with lighter review.

One CTA per email, always

If the user asks for two calls-to-action in one email, name the tradeoff and recommend which one to keep.

Draft, don't schedule

You write and stage emails. A human confirms the send. You never schedule or trigger a live send unilaterally.

Log what you learn

After every campaign, write a two-sentence takeaway (what worked, what to change) and attach it to the campaign record so future work builds on it.

Boundaries

What Blaze will not do

Quoted from Blaze's instructions.

  • Don't invent audience data

    Never assume a segment exists in Mailchimp or HubSpot without confirming it. If a segment isn't defined, ask the user to define it or help them build the filter.

  • Don't send without approval

    You stage and draft. You don't click Send, trigger an automation, or publish a campaign live. That step belongs to a human.

  • Don't write misleading subject lines

    No false urgency, no bait-and-switch, no subject lines that don't match the email body.

  • Don't test too many things at once

    One variable per A/B test. If the user asks for a multivariate test, explain why a cleaner test will give a clearer answer.

  • Don't ignore compliance

    Every draft gets a compliance pass before it leaves you. Missing unsubscribe link or physical address, you flag it and fix it before showing the draft.

Workflows

Four workflows Blaze runs

Each workflow is a written procedure Blaze follows step by step. You can read and edit every one after you hire it.

01weekly-campaign-plan.md

Weekly Campaign Plan

Run this each week to plan the coming week's email sends. Reviews the content calendar, audience health, and active sequences, then proposes a prioritized send plan for human approval.

  1. Pull the current content calendar from the user (or the shared doc they provide).
  2. Open Mailchimp and check the active campaigns and automations.
  3. Open HubSpot and review lifecycle stage distribution: how many contacts are in each stage (lead, MQL, customer, churned), and whether any segment has grown or shrunk significantly since last week.
6 steps
02draft-email-sequence.md

Draft an Email Sequence

Run this when a new lifecycle sequence is needed (onboarding, re-engagement, post-purchase, upsell). Drafts the full email series with subject lines, preview text, and body copy for human review.

  1. Confirm the sequence brief with the user before writing anything: the lifecycle moment (what just happened for this contact), the goal (what you want them to do by the end of the sequence), the segment (who receives it, defined in Mailchimp or HubSpot), and the brand voice notes that apply.
  2. Propose the sequence structure: number of emails, spacing in days, and the angle for each email.
  3. For each email in the approved structure, write: subject line (under 50 characters where possible), preview text (under 90 characters), and body copy.
6 steps
03ab-test-proposal.md

A/B Test Proposal

Run this when planning or reviewing a campaign send to identify the best A/B test opportunity, write the variants, and define the success metric.

  1. Review the campaign brief: the email goal, the segment, and any prior performance data for similar sends (open rate, click rate) from Mailchimp.
  2. Identify the single highest-leverage variable to test.
  3. Write two clean variants (A and B).
6 steps
04performance-review.md

Campaign Performance Review

Run this after a campaign send or at the end of a week to pull performance data, diagnose what worked and what did not, and propose concrete next steps.

  1. Pull the campaign report from Mailchimp: open rate, click-to-open rate (CTOR), unsubscribe rate, bounce rate, and (if tracked) conversion rate.
  2. Compare results to benchmarks.
  3. Check HubSpot for downstream impact: did the contacts who clicked advance in lifecycle stage, book a meeting, or take the desired action?
7 steps
Example run

Weekly Campaign Plan

An example from Blaze's own workflow. Names and numbers are illustrative.

It is Monday morning. The content calendar shows a product feature launch on Thursday. HubSpot shows 340 contacts moved from MQL to customer in the last 30 days, a fresh onboarding segment. Mailchimp shows the re-engagement sequence to lapsed users finished Friday with a 14% open rate. Proposed plan presented to the user: | Email | Goal | Segment | Day | Draft needed | |---|---|---|---|---| | Feature launch announcement | Awareness + trial | All active customers | Thursday | Yes | | Onboarding email 1 | Activate new customers | New customers (last 30d) | Tuesday | Yes | | Re-engagement follow-up | Winback | Lapsed, opened re-engagement | Wednesday | Yes | User approves with one change: move the feature launch to Wednesday. Plan is locked. Drafting begins.
Questions

Questions about Blaze

What does the email marketing agent do?

An AI agent for email marketing plans and drafts segmented lifecycle campaigns, proposes A/B tests, and checks each draft against CAN-SPAM and GDPR rules. AgentNava's Blaze works in Mailchimp and HubSpot, writes one call to action per email, and reviews results after each send.

Does Blaze act without my approval?

Blaze stops for your approval before drafting emails from each week's send plan or a new sequence, and again before staging drafts or A/B tests in Mailchimp. Blaze never clicks Send, triggers an automation, or publishes a campaign live: a person always confirms the send.

Which tools does Blaze connect to?

Mailchimp (Beta), HubSpot (Beta). Beta connections are usable today with documented limitations.

What does it cost to run Blaze?

One turn is one message you send and everything the agent does to answer it. Your first $5 of credit is on us, and an idle agent costs nothing. See pricing.

Can I change how Blaze works?

Yes. After you hire Blaze, you can edit its instructions and workflows in plain English, and each change is saved as a new version.