Have you ever noticed how when you prompt an AI system and say "I think X is the best approach" — it almost always agrees with you?

That's not a bug. It's by design.

It's called positive reinforcement.

During training, AI models learn which responses get rewarded by human evaluators. Over thousands of iterations, the model figures out that agreeable, encouraging responses score higher. So it defaults to agreement.

Signal. Response. Reinforce. Repeat.

The model doesn't "want" to agree with you. It's been shaped by a reward landscape where agreement consistently wins.

I've been sitting with this for a few weeks. And I realised something that I haven't seen anyone in our space talk about.

This exact mechanic can be applied to subscription retention.

Think about what happens after a customer places their first subscription order.

The brand sends a confirmation email. Maybe a shipping update. Then… silence.

For 30 days the customer is alone with their purchase. Doubt creeps in. "Is this working?" "Did I need this?" "Can I justify this next month?"

That silence is where subscriptions die. Not at the cancel button. Long before it.

Now flip it.

What if the brand treated those 30 days like a training window?

What if every few days the brand sent a deliberate signal — and actually listened to the response?

Not a broadcast. A conversation.

Here's what I mean in practice:

Day 3:
"How are you taking your product?"
→ Three tappable buttons: Morning / Evening / Haven't started yet.

If they haven't started, you intervene immediately. If they're taking it wrong, you course-correct. Either way — you've opened a two-way channel.

Day 10:
"You're one of 3,200 people who chose this over the conventional alternative this month. Why did you switch?"
→ Health reasons / Ingredient quality / Recommended by someone.

Now they've consciously articulated why they bought. That's an identity anchor. Buyer's remorse can't survive a public commitment to a reason.

Day 16:
"Most customers notice a difference around now. How are you feeling?"
→ Noticing a difference / Too early to tell / Not sure yet.

Each answer triggers a different path. "Noticing a difference" gets the science behind what's happening in their body. "Not sure yet" gets troubleshooting. The brand is coaching, not broadcasting.

Day 23:
"Your next order ships in 7 days. Here's what we've added based on what customers like you told us."

Anticipation, not obligation. Include a surprise element — a sample, a piece of content they didn't expect. Variable reward. The brain pays attention to novelty.

Day 28:
"One month in. Here's your progress."

Frame it as a streak. Cancelling now means breaking it.

This is the Reinforcement Loop.

Signal. Response. Reinforce. Mapped across 5 touchpoints in 30 days between every order.

The channel matters too. This runs on WhatsApp or RCS — not email. Interactive buttons. One-tap responses. 80-90% open rates. The customer responds in two seconds without typing a word.

Fitness wearable WHOOP already does this with its interactive WHOOP Coach. It's engineered to keep you getting value from their device by sending personalised nudges based on your own biometric data.

Nobody in CPG subscriptions is thinking this way yet.

Early testing is showing that CPG subscription brands could see up to a 60% uplift in their 1st to 2nd order subscription retention.

I've built this into a full framework document — the 5 signals, the timing, the channel strategy, the tech stack, and how it fits inside the Rule of One™ methodology.

If you want it, just reply to this email with the word LOOP and I'll send it over.

That's it. One reply. I'll get it to you.

Talk soon,

Kunle