Moats in the Age of AI: Why Your Real Advantage Isn’t Your Model, It’s Your Ability to Learn
A practical guide to building defensibility in an AI-first world
After seeing and commenting on two posts on LinkedIn (Jessie Sheff [Principal at Insight Partners] and Phil Edmondson-Jones [Partner at Oxx]) about the ongoing ‘moats in AI’ topic, I thought I’d add my take. This is my first Newsletter post. So, if you like it, please subscribe below and I’ll write more.
Note: Product Leaders is intended to be a newsletter for real product operators who want to get product right in the real world. I’ve noticed an increase in fluff, noise and non-actionable articles on LinkedIn recently. So rather than moan about it, my goal is to cut through the noise to collate some grounded, commercially sharp guidance for product leaders, founders and C-suite teams. No jargon. No hype. Just practical insights on product strategy, execution and building teams that learn fast and deliver with intent. If you care about this rather than chasing the next framework, this is for you.
0. Most conversations about moats in AI still orbit the same clichés.
Data. Models. Scale. Flywheels. All useful, but increasingly insufficient.
Because the truth is uncomfortable for both startups and incumbents: in an AI-first world, the strongest moats are composite, not singular. And the only thing I believe that compounds faster than technology is learning.
1. The old moats haven’t disappeared; they’ve become interdependent
In the past, you could defend a market with one strong edge. Today, the advantage sits at the intersection of three forces:
1. Proprietary data - Not just volume, but structure, meaning and history.
2. Organisational design and behaviour - How fast your teams turn insight into action.
3. Product experience - The quality of your UX, your workflows, your nudges, your activation paths.
Individually they help. Combined, they compound. This is what I meant by “composite moats”.
The defensive wall isn’t a single asset; it’s the connection between them. Think of it as a flywheel of strategic learning rather than strategic assets.
2. Why AI makes adaptation the real moat
AI has levelled the playing field in ways people underestimate:
Anyone can spin up a model
Anyone can plug into an API.
Anyone can deploy an LLM-powered feature and stick “AI” in the roadmap.
So the question stops being “Can you build it?” and becomes:
“How quickly can you adapt your product and organisation as the world shifts under your feet?”
Incumbents don’t lose because they lack tech.
They lose because they can’t rewire how they work.
The pain is cultural, not technical.
If your release cycles still depend on committees, your moat has already eroded. *insert empowered teams here*[sic] (This is a whole other newsletter waiting to be written about the millions of empowered teams out there with no clear mandate, no actual authority, lacking the appropriate skills or resource and many other dysfunctions yet are really only empowered in name when you get recruited…)
3. The only durable moat: your rate of learning
I’ve held a simple belief for years:
Your learning cadence beats your planning cadence.
Every time!
It’s why I use my own Winning Team Formula:
Work (transparently + small + meaningfully + swiftly) + Inspect/Adapt = high-cadence learning, compounding value, real accountability.
I’m pretty sure this is inspired by something John Cutler shared many years ago, but I can’t find the link. I’ll update if I can find it.
It’s intentionally unsexy.
No jargon. No mythical frameworks - although like any good product person and agile coach, I do, have and will continue to use what really matters as needed.
Just the habits that allow teams to move quickly without creating a mess.
But – and this is the uncomfortable part –it only works if Product is empowered to align the organisation around learning.
If Product doesn’t have the mandate to activate the full depth of the company’s data, talent and IP, the loop breaks. You get “activity”, not progress. You ship, but you don’t learn. You get asked ‘when is that thing going to be done?’.
4. So what does this mean for operators trying to build moats today?
Here are hopefully the practical signals that your organisation is building real defensibility and if not, you could consider:
A. You have a structured learning engine
Not just ad-hoc insights, but a repeatable loop:
Clear hypotheses that are documented and tested
Fast experiments and not too many at a time you don’t know what really had an impact
Honest read-outs with the right audience to make a different
The whole organisation feeding the same learning backlog
B. You treat data like an operating system, not a warehouse
It’s accessible.
It drives decisions.
People know how to get what they need from it and time has been invested in training them to actually use it as they need
Your AI features don’t exist in isolation; they make the whole product sharper.
C. Your product surface area is designed to compound
Every workflow, every nudge, every touchpoint helps you learn something useful about behaviour. (Massive Posthog fan here for this sort of thing)
This is where UX becomes a moat, not just a cosmetic layer.
D. You’ve built organisational muscle around adaptation
Teams make small bets. Lots of them. And you actually realise a lot of them didn’t work. And that’s ok.
Leaders remove drag, not add process. They actually get out of the way.
You ship, measure, adjust and repeat. Simple really…note you don’t measure, adjust, repeat, then ship. The order matters!
E. Your culture rewards truth over comfort
If your metrics tell you something painful, people aren’t punished for saying it aloud.
Moats rot when people protect their slide decks instead of their users.
5. The uncomfortable but liberating conclusion
AI has not killed moats.
I think it has raised the bar on what counts as one.
So this is my thesis (or is it a princple, or mini-manifesto, or mental-model [another impending article]):
Your product wins when your organisation learns faster than your competition.
When you can turn data + behaviour + UX into a system that evolves continuously.
When teams know how to operate in ambiguity because they aren’t waiting for permission.
As true as I will always order whatever variation of Pepperoni and Nduja if its on the pizza menu…Tech will keep accelerating. LLMs will keep closing the gap between idea and implementation. But the companies that pull ahead will be the ones that can answer one simple question:
How quickly can you learn the thing your competitor hasn’t realised yet?
That is the real moat.
And it’s entirely within your control*
*unless you don’t actually have control over your product - then that is a whole volume of newsletters waiting to be written!