Dormant-Customer Re-engagement Agent
Turn dormant customer records into a measured three-touch revenue campaign
- Difficulty
- Moderate
- Time to result
- ~days to results
- Steps
- 7
- Confidence
- 99%
The re-engagement agent searches connected CRM, billing, and database systems for past customers or leads who have gone quiet but remain plausible buyers. It prioritizes high-potential records, then creates a relevant offer and a personalized three-message sequence delivered on day zero, day three, and day six. The agent can initially prepare a review document or drafts rather than send automatically. Once approved, it uses its dedicated inbox to execute the sequence, monitors responses, and brings a human into live opportunities. Performance data such as replies or open rates becomes feedback for the next bounded batch. This turns a dormant warm audience into a recurring, measurable pipeline process rather than an occasional manual campaign.
Origin
Extracted from Marketing Against The Grain as the concrete first AI employee built by Nick Velescu with Hermes.
Core principles
- 01Past customers and leads are warmer than cold prospects
- 02Targeting and offer design precede outreach
- 03Personalization should use available customer context
- 04Replies should trigger human involvement
- 05Campaign performance should drive the next iteration
How to run it
- 1
Connect Customer Sources
Give the agent authorized access to the CRM, billing platform, database, or other systems containing customer history.
Pro tip Combine transaction and engagement data when both are available.
Watch out Do not send outreach until contact permissions and record quality are understood.
- 2
Select Re-engagement Candidates
Find former customers and warm leads who have not engaged recently, then rank those most likely to benefit from renewed contact.
Pro tip Use prior purchases, fit, and elapsed time as targeting signals.
Watch out Do not treat every dormant record as equally valuable.
- 3
Craft the Offer
Develop a reason for the recipient to re-engage with the product or service now. Tie it to known customer context where possible.
Pro tip Make the offer specific enough for the agent to personalize consistently.
Watch out A generic check-in without a useful offer is unlikely to reactivate demand.
- 4
Build the Sequence
Create three personalized emails with distinct purposes and schedule them for day zero, day three, and day six.
Pro tip Use later messages to add value or address likely hesitation rather than merely repeat the first email.
Watch out Avoid identical follow-ups that add no new reason to respond.
- 5
Review Before Sending
Have the agent produce a customer list, proposed messages, and review document for approval during the initial rollout.
Pro tip Start in draft mode if the team is not ready for autonomous sending.
Watch out Do not approve a campaign without sampling both targeting and copy.
- 6
Monitor and Escalate Replies
Track campaign responses and bring a human into the thread when a recipient shows interest or requires judgment.
Pro tip Define explicit handoff conditions before launch.
Watch out Do not let valuable replies sit unattended in the agent's inbox.
- 7
Measure and Improve
Review results after a bounded volume, ask the agent what could improve, and apply the approved changes to the next batch.
Pro tip Compare changes on a fixed next batch rather than changing the sequence continuously.
Watch out Do not scale poor targeting or messaging before diagnosing the result.
In the wild
The demonstrated agent searches connected Stripe and customer systems, identifies promising former customers, proposes who should receive outreach, drafts a three-email sequence, and presents a review link. After approval, it sends from its own Agent Mail inbox and asks for human involvement when someone replies.
→ A neglected customer list becomes an automated but supervised source of renewed sales conversations.
Common mistakes
Automating Before Reviewing
A first campaign should expose its target list and messages for review or operate in draft mode until quality is established.
Ignoring the Handoff
The agent should pull a human into promising reply threads rather than attempting every sales conversation alone.
Scaling Without a Checkpoint
Sending indefinitely without a bounded performance review can amplify poor targeting or weak copy.
Is it for you?
Best for
Businesses with a meaningful backlog of past customers or warm leads that have not been contacted recently.
Not ideal for
Businesses without permission to contact the audience or without reliable customer records and offer context.
From the transcript
“what I want to build today with you is an AI employee that engages these past customers, past leads, makes an email sequence, warms them…”
“Here's the personalization it's gonna do. Here's the timing of it. Here's the offer that it's gonna create to re-engage these customers. Uh yeah, day…”
“Do the first thousand and then send me an email about the performance of that first thousand and how it went. And what we could…”
From the episode
Build An Army Of AI Employees To Run Your Business (With Hermes AI)