Episode 86 - Retention AI Maturity Model + Build In Public Progress Update + Course Announcement

Welcome to episode 86 of The Retention Blueprint.

Most retention teams are still operating either on rules-based logic or, at best, predictive models.   

The Retention AI Maturity Model is a five-level framework for understanding where you are and what it actually takes to move up the AI maturity curve.

Level 1 is rules-based logic. Send an email if someone hasn't logged in for 7 days. Flag a customer as at risk if they don’t complete onboarding. Generative AI co-pilots live here; they speed up the execution of the old paradigm. 

Level 2 introduces predictive targeting: churn scores, propensity models. Useful, but still reactive by nature, and the signals are always late. Automated execution and ingestion of retention journeys using tools like Claude code live here. They are talking months off execution (see the build-in public article below), but still live in the old paradigm. 

Level 3 is where the real shift happens. Embeddings. Instead of rules or scores, you're building a behavioural fingerprint for every customer, detecting drift weeks before a standard churn or growth signal appears. 

Level 4 introduces assisted agents: AI that recommends the right intervention, with human oversight built in. High-value customers, regulatory edge cases, complex pricing: humans stay in the loop where it matters.

Level 5 is autonomous agents. Moment-specific, self-improving, operating within pre-approved guardrails. A fleet of agents that scale your retention execution while protecting margin.

The biggest mistake? Trying to skip from Level 1 or 2 to Level 5. Doing so just creates autonomous agents on top of old workflows and carries multiple risks.

👉 Read the full Retention AI Maturity framework

CRM Marketing 2.0 Course Announcement 

After 17 months, my CRM Marketing 2.0 is closing at the end of April. Across v1 and v2, over 1,500 students have taken the course, and the feedback has been brilliant. I'm moving on to build something bigger, but if you've been sitting on the fence, this is your last chance to get it at $149 before it closes for good.

👉Find out more about the course 

Build in public progress update: Subscription App Client 

This is part 2 of my build-in-public project.

To recap and for those new to this, for a subscription App client, I am using Obsidian (containing 26 years of retention & CRM IP), Claude Skill, Nano Banana and Claude Code to write all copy, produce all images and automatically ingest an entire lifecycle program across 7 lifestages and 3 channels into Airship. Taking a 3-4mth project to under 30 days while maintaining high quality execution. 

So what's happened so far? 

1. Obsidian knowledge base built! It was a fairly straightforward process of cutting and pasting all my relevant CRM Marketing IP into Obsidian and creating linkages between ideas. It was time-consuming to do, but it's reusable. I call this my Retention AI second brain. 

2. Next, I fed 40+ MD files, from my Retention AI second brain, into Claude to write a Skill for subscription apps. This Claude skill includes everything I know about behavioural science, lifecycle marketing, CRM marketing UX, moments of truth, behavioural drift, time to next meaningful action and more, but made specifically relevant to subscription apps. 

3. I then wrote a 15-slide strategy deck for the client. This deck includes 10 principles written specifically for this client, such as how to adhere to the personalisation paradox in their context and why building the relationship on mutual trust is critical for this business. As well as rules about time to next meaningful action and a high-level communications framework. Everything in the deck is entirely bespoke to the client’s specific market and challenge, but again fully aligned to what I know works and doesn’t.  

4. Next, I combined the Claude Skill skill for subscription apps with the client-specific strategy deck and asked Claude to create a communications framework. It took about 10 minutes for Claude to create this. Honestly, I was extremely impressed. I made 3 edits, the client made 3 more. This framework includes life stages, journey stage, trigger, communication name, timing, channel, audience criteria, personalisation rules, data points, business objectives, customer mindsets, identity stages, and reinforcement triggers. Massive. 61 hyper personalised communications across email, in-app and push across account creation, free-trial, first 90 days, in-life active, in-life loyal, cancelling and winback lifestages.

5. Next, I asked Claude to create the copy and image briefs. The client and I then spent an afternoon locked in a room together reviewing all the copy line by line and every image brief in detail. This took about 4 hours, and we had about 10-15 edits, plus 3-4 new ideas. 

6. Then I asked Claude to create a technical spec to ingest everything into Airship via the API, including all events to be tagged in the app and steps required by the developer. This document is exceptional, 23 pages of detail. 

The next step is for Claude Code to write the script and ingest it all into Airship. We have a developer helping, but the developer's role is app instrumentation and deployment. Claude Code handles all script writing, API integration, and the content upload pipeline. 

Once this is done, all journeys and templates will be automatically built, saving masses of time. This is what we are doing next, I’ll update in the next episode on where we got to. 

In the meantime, if you want to learn more about my AI second brain, check out this post. 

Until next time, 

Tom