Every medium business already owns the cheapest revenue available: the customers it has already paid to acquire. Re-engaging an existing customer costs up to 25x less than winning a new one. Yet most companies stop talking to a buyer after one purchase and go back to prospecting cold. The revenue is not missing from the data. It is missing from the calendar.
The 90-day silence
A customer who bought from you a year ago and heard nothing since is not a lost cause. They are a list item nobody has time to work. The typical pattern is a burst of post-purchase emails, then silence, then a discount blast to everyone that trains the best customers to wait for the next sale. None of that is a reactivation program. It is noise.
The fix is a standing operating procedure, not a campaign: any contact who has not opened an email in 90 days receives one short message with a single question and a single link. Non-responders are removed to protect deliverability. The rule sounds small. Executed every week, it is the highest-ROI automation most small and medium businesses can install.
Why this needs both automation and AI
The cadence belongs to automation. Purchase-history triggers, inactivity windows, and removal rules run whether or not anyone remembers. The message belongs to Gen AI. Tone matched to what the customer bought and when, a reactivation draft, a win-back draft, a cross-sell draft, all written for human review or autonomous send inside rules you set.
The multiplier is capacity. A sales or marketing generalist stops manually filtering dormant lists and writing one-off emails, and instead supervises a system that works the database every week. AI customer-marketing agents are now positioned as doing the work of three roles for a fraction of the cost of one junior hire. For a medium business, that is the difference between having a reactivation program and not having one.
Segmentation by engagement, not by purchase date
The old habit segments by what people bought. The useful segmentation is whether they still open anything. An engaged buyer who purchased once is a cross-sell candidate. A dormant buyer who purchased twice is a win-back candidate. Someone who has opened nothing in 90 days is a reactivation candidate, and after one clean attempt, a suppression candidate. Automation enforces those buckets; AI writes to each bucket in the right voice.
Run the experiment in 30 days
- Pull every contact with no email open in 90 days.
- Split them by last purchase: never bought, bought once, bought twice or more.
- Send one reactivation message per segment: one question, one link, no discount unless the segment is proven discount-driven.
- Measure reactivation rate, not open rate. A reactivated buyer who purchases again is the only result that counts.
- Remove non-responders. Protect the list.
The measurement question matters. Open rates flatter vanity. A reactivation rate of even a few percent on a dormant base is pipeline you did not pay to acquire.
The design rule
Automation owns the trigger, the rule, the removal, and the audit trail. Gen AI owns the drafting and the personalization. Humans own the offer that touches margin. That boundary is what keeps reactivation a system instead of a guilty quarterly blast.
Your cheapest customer is the one you already paid to acquire. When did you last speak to the one who bought from you a year ago?