Shrinking the Learning Loop

At Mostly Working this morning, James Storer led a fascinating walkthrough of how Monzo are building AI infrastructure to speed up prototyping.

Reflecting afterwards, I realised how rarely he used the word ‘design’. Instead, the emphasis was on product-wide outcomes like faster learning loops and higher-quality prototypes. It felt refreshingly disconnected from shipping to production faster, the dominant narrative for this stuff. Even if that’s where it ends up, it’s not the only value along the way.

A few other takeaways, all of them about the same thing: tightening the loop on deciding what to build.

1. Internal prototype tooling pull was the real signal.

Their React Native prototyping app reached 120 users internally, finding its own product-market fit inside the company. What struck me was the spread across design, engineering and product. Building that cross-functional ‘coalition’ felt like a clear theme in broad adoption, not just individual disciplines.

2. Deeper personalisation drives more insightful research.

The accuracy of content and data in a fintech product plays a big role in how real it feels to participants, which lifts the signal you get back. Monzo have built neat tools to generate content that makes a prototype feel believable enough to sharpen those sessions. In other words, better context in, better learning out.

3. Custom apps lower the friction to getting set up.

James’ team replaced their ‘get started’ docs with a native Mac app that walks new joiners through setting up their machine. Basically a nice UI layer over terminal commands, guiding people through ‘copy + paste this here’. Docs go stale fast when the tool stack evolves this quickly, but keeping it all in one place means onboarding can be optimised like any product. As the saying goes, every product is two: the product, and onboarding to use it.

4. AI capability compounds over time, especially when it’s shared.

One person prototyping faster doesn’t change much. The shift is turning that into something the whole org can do: AI buddies, lunch-and-learns, and a small ops group working alongside the platform and infrastructure teams. Each new person or discipline adds capability that compounds on the last.

+++

James was refreshingly honest. The strongest wedge has been shrinking the learning loop to explore faster and build higher conviction in where product teams are headed. One example stood out: they caught a direction that wasn’t working early enough to save three months building down the wrong path. Shipping to production still needs time to figure out, and that’s probably a better problem to tackle once standards, tools and workflows mature. Even so, there’s huge value to deliver in the meantime, while building the organisational muscle and infrastructure to do this properly. Not (necessarily) shipping faster, but learning faster what’s worth shipping at all.

Oh and James’ AI image generation game was on point. I hope you like my fancy sweater.