Episode 94 - Bellingham, Uber and Agentic CRM

Welcome to Episode 94 of the Retention Blueprint. 

Something happened last week I want to tell you about.

I got on a call with one of the people who built the messaging engine at Uber.

He'd seen my content on agentic CRM, and he reached out.

Everything I've been writing about where CRM is heading, one-to-one, in real time, decided by agents, governed so a board can actually sign it off, he built at Uber.

It's the same architecture I build with clients now, just at a scale most of us will never touch.

We both agreed, it's just how you do CRM properly.

So this week, let me show you how Netflix and Uber really run CRM. 

Because it maps exactly onto what I call CRM 3.0.

Netflix

Netflix runs what they call a "slow-fast" architecture.

A slow planner: an offline agent that builds a weekly pacing plan for every single user. It weighs the value of messaging you against the costs: fatigue, opt-outs, and even the risk of reminding you to cancel.

A fast executor: a real-time engine that, at each send opportunity, checks the plan and picks the best message and artwork in the moment.

The key here is that the content is picked; it is not written by the agent at send time. 

The decision is separated from the send. The frequency is governed. And the content isn't baked in; the agent carries pointers to it.

Uber

Uber does the same thing, a different way.

Every message is built from slots, subject, banner, body. Models predict what each individual is most likely to respond to, then assemble the best combination, per person.

Then there's a real-time layer. Change your location, and the offers in your queue update instantly.

And again, this is the bit that matters most: the AI never free-writes to the customer. It generates and scores variants behind a rule engine, then selects from what's already approved. It selects. It doesn't author.

Neither Netflix nor Uber uses a visual journey builder.

No canvas. No fixed flows.

They build large menus of approved content and let agents read real-time signals to choose the right thing for each person, every time.

That is CRM 3.0.

And it isn't only for companies with hundreds of millions of customers. Most enterprises already own most of the pieces. The work is in defining the action space, building the governance, and creating the content.

Scaling content production 

Most brands get stuck in content production.

And most people make the mistake of assuming that the AI will free-write the content.

But if you do that, you're just asking for risk, and it's not how Netflix or Uber do it.

This is where my AI second brain comes in.

I've encoded 26 years of retention work, 100+ case studies and frameworks, into a knowledge base AI can read. 

So what I do is integrate my AI second brain with Claude and the clients' strategy. 

So when Claude drafts, it's grounded in proven retention principles, behavioural science, and the client's own IP and context. 

Then we (humans) read it all. 

And only once it's approved does it feed the engine. 

On a recent World Cup project, we ingested player and team history and produced over 3,200 communications.

Thousands of one-to-one variants: tuned to a single match or a whole run of them, with the right team and player history, the right offer, and content that fits each audience.

A variant might lean on Bellingham's two goals against Norway to build anticipation for tonight's fixture. All grounded, all pre-approved.

Everything is signed off before it is ever sends.

Which brings me to next week

I walked through the full architecture in my session last month. If you missed it, or want to go back over it, it's here:

👉 https://www.youtube.com/watch?v=5DdNogjS7zc

And next week I'm taking a small group deeper.

Thursday the 23rd. One hour. Invite only. Up to 10 brands, each bringing one or two real questions, worked through in detail.

Most of the spots are now gone. This is the last call for the few that are left.

To join, you'll need to be Head of CRM or more senior (CCO, CMO or equivalent), working in-house at a streaming, gambling, sports, subscription app or publishing brand.

If that's you, just reply to this email and tell me you'd like in. 

I'll send you the details.

Until next time,

Tom

P.S. What did you think of this episode?

😎 It was great!

🤷 It was ok

🫤 Could be better

Retention Support: When you are ready, reach out to discuss consultancy, mentorship, fractional support, training or help in building your agentic CRM roadmap.