Human-centered design systems and product experiences
Core services
01UI/UX Design02Web Development03Software Development042D/3D Experiences05App Development06AI Agents & Automation07Website DesignTechnical services
Frontend Engineering
React DevelopmentNext.js DevelopmentWeb ApplicationsEnterprise DashboardsSaaS PlatformsBackend & Infrastructure
API DevelopmentCloud InfrastructureDevOps & CI/CDBackend SystemsCustom SoftwareIndustries we serve
Technology & Software
AI Customer Reactivation Agent Development
Quiet customers, invited back
An agent for the customers who stopped coming.
An agent for the customers who stopped coming. It finds who has lapsed, removes everyone who may not be contacted, works out from their history what would be relevant, and prepares the message for your team to review before anything is sent.
Every business has a list of people who bought once or stopped a year ago. Occasionally someone exports it and sends everyone the same "we miss you" email. It reaches people who complained, people who asked not to be contacted and people who moved away. The few who might return get a message that shows nobody looked at what they bought.
This page describes a workflow we propose to build as a custom project. It does not exist as software yet. It would reuse the screening, drafting and reply handling that run in our outreach preparation demonstration on test data. Reading purchase history, segmenting lapsed customers and choosing a relevant reason to write are not built.
Where the Time Goes Now
One message for everyone
The customer who bought a pram receives an offer for the same pram.
Writing to people who left unhappy
An open complaint is followed by a cheerful marketing email.
Old consent, new channel
Someone who agreed to email is messaged on WhatsApp.
No definition of lapsed
Weekly buyers and yearly buyers are measured against the same gap.
Lists nobody cleans
Closed accounts and duplicate records are contacted again.
Activity mistaken for return
Opens are reported as reactivated customers.
How One Request Moves Through
An example of one case, in 6 steps. The same steps play in the picture at the top of this page.
Segment defined
Your team defines lapsed as no order in twice the customer's usual interval.
List built
The agent finds the customers who match and shows how each was selected.
List screened
People with opt-outs, open complaints or closed accounts are removed, with the count per reason.
Reason chosen
A customer who bought a coffee machine a year ago is matched with the descaling kit it needs.
Drafts reviewed
Your marketer reads a sample, edits the template and approves the batch.
Replies handled
Interest is passed on, one complaint goes to service, two opt-outs are recorded.
Six jobs it carries through
Lapse detection
Finds customers who have gone quiet for them.
- Uses each customer's own buying rhythm
- Applies the definition you set per segment
- Excludes recent contacts
- Reports how the list was built
Eligibility screening
Removes everyone who should not be contacted.
- Opt-outs and channel consent
- Open complaints or disputes
- Closed, merged or duplicate records
- Your own exclusion rules
Relevance
Works out what is worth saying.
- What they bought and when
- What usually follows that purchase
- What has changed since they left
- Nothing when there is no good reason
Draft for review
Prepares the message.
- Uses your approved templates and tone
- Refers to their real history
- Includes only offers you approved
- Waits for a named approval
Reply handling
Handles what comes back.
- Interest goes to sales or service
- A complaint goes to your service team
- An opt-out is recorded at once
- Everything else is left for a person
Honest reporting
Reports what happened.
- Customers selected and excluded, with reasons
- Messages sent, by channel
- Replies by kind
- Orders placed afterwards, listed separately
What We Need to Scope This
- The process as it runs today, in a few lines.
- The systems it touches, for example customer data, order history, consent, service history.
- Rough monthly volume, and who approves what.
Fewer Messages, Each With a Reason
The agent's first job is to remove people from the list: everyone without consent for the channel, everyone with an open complaint, every closed or duplicate record. What remains gets a message only when the customer's own history gives a reason to write.
Sending is done through your platform after a named person approved the batch. What the agent reports is what it did, with orders listed separately from messages.
- Your definition of lapsed
- Consent checked per channel
- No contact during an open complaint
- A reason from the customer's history
- Named approval before sending
- Opt-outs remembered
Where It Connects
The systems named here are the kind this agent works with. Which ones it connects to in your case is settled in the scope.
Customer data
- CRM
- shop system
- loyalty programme
Order history
- Purchases, dates and categories
Consent
- Opt-in and opt-out records per channel
Service history
- Open complaints and disputes
Sending
- Your email or messaging platform, used with your approval
Catalogue
- What is relevant to follow a past purchase
Asked Before the First Call
How is this different from an email campaign tool?
A campaign tool sends what you load into it. This agent does the work before that: deciding who is genuinely lapsed, who may not be contacted and what would be relevant to each person. It prepares a smaller, better list and hands it to your sending tool.
How is this different from the renewal and retention agent?
Renewal works on customers who are still with you and have a date coming up. Reactivation works on customers who already left or went quiet. The first prevents a loss, the second tries to reverse one.
Can it tell us how many customers it brought back?
It can list the orders placed after its messages, separately from the messages themselves. Whether a message caused an order is a question it does not answer by itself. A fair measure needs a comparison group that was not contacted, which we can set up with you.
What exists today?
Not as software yet. This page describes a proposed workflow. It would reuse the screening, drafting and reply handling that run in our outreach preparation demonstration on test data. Reading purchase history, segmenting lapsed customers and choosing a relevant reason to write are not built. A project starts by mapping your process, then proving the first steps on test data before anything touches a live system.
Start by Defining Lapsed
Bring your order history. The first useful result is usually the list of people who should never have been on the old reactivation list.