Moats for the decade when building is cheap
Seven durable advantages for businesses that throw off cash, read through Charlie Munger’s latticework of mental models.
Buffett gave it the name. Munger spent sixty years teaching people to tell a real moat from a painted one.
For most of software’s history, the build itself was the moat. Shipping something good took a team, a year and a few million dollars, and that cost kept the field thin.
That cost is now close to zero. A competent operator with a good AI model can clone a product built on features over a weekend and fill a content library in an afternoon.
That leaves the question Munger would ask first. Once building costs nothing, the only advantages left are the ones that stay scarce.
“It was hard to build” stops protecting anyone. The seven moats below come from the businesses I run and the ones I watch, and they decide what I build and what I buy.
Invert first: what erodes
Munger borrowed a habit from the mathematician Carl Jacobi: invert, always invert. Before naming what will last, name what is already dying.
- SaaS sold on features, and thin AI wrappersAnything a weekend of prompting can rebuild was never defended.
- Pricing by the seatAgents shrink seat counts, and revenue tied to headcount shrinks with them.
- SEO content and info productsSupply is becoming infinite, and price follows supply to the floor.
- Generic agencies selling labor hoursThe machine works faster than the clock, and billing the clock loses.
What stays scarce, ranked by fit for a lean operator
Each moat below sits next to the Munger model that explains why it holds.
-
Owned distribution and trust
An audience that reaches you directly through email, a podcast or a community is the scarcest asset when everyone can ship product. It is the moat most worth compounding. An audience rented from an algorithm doesn’t count.
-
Proprietary data that accumulates with use
This is data the business produces by running: outcomes, benchmarks, transaction histories and a record of what worked for whom. Scraped data is open to anyone. This kind improves the product every month in a way a clone can’t copy.
-
Being the system of record
If the customer’s money, records or daily workflow lives in your product, switching is painful. Vertical software for unglamorous niches, especially with payments embedded, holds up far better than horizontal tools.
-
Accountability for outcomes
AI does the work, and someone still has to own the result. Selling finished work with a guarantee (books closed, leads recovered, claims filed) earns software margins inside a service relationship. The moat is trust plus operational skill.
-
Curated access to people
Vetted peer groups, niche marketplaces and deal networks can’t be generated by AI. The vetting is the product.
-
Physical and local presence
Distribution, home services, and anything with trucks, inventory or licenses qualify. These businesses are boring and throw off cash, and AI can modernize their back offices.
-
Regulatory permission
Licenses, compliance approvals and certifications are the strongest moat on this list and the slowest to get.
The tension: your circle of competence cuts both ways
The filters I operate by (short sales cycles, light regulation, minimal headcount) keep me inside my circle of competence. They also screen out moats 6 and 7, along with part of moat 3.
Whatever is easy for me to enter is easy for everyone else too.
The workable answer is to stack the first few moats until they reinforce each other. Munger had a name for several forces pushing the same direction at once: the lollapalooza effect. The combined result runs far larger than any single force produces alone.
Audience into a vertical product into a proprietary data loop
Use distribution to enter a niche cheaply, become the place where the work happens, and let usage data widen the gap every quarter.
Services priced on outcomes and delivered by AI in boring B2B niches
Recurring revenue, a short sales cycle and a small team, with incentives pointed at the result the customer wants.
A minority stake or the growth layer in a physical business
Own a piece of the trucks and licenses, or run the demand engine behind them, without taking on the job of operating one.
Avoid stupidity before chasing brilliance
Munger credited much of Berkshire’s record to consistently avoiding dumb decisions instead of hunting for genius ones. Applied to the next decade, that turns into a short set of refusals.
I won’t build on an audience a platform can take back. Pricing by the seat makes no sense when customers are cutting seats, and billing hours makes less sense every month the work gets faster.
The last refusal matters most. Any advantage that exists only because something was hard to build last year is already gone, even if the revenue hasn’t noticed.
Invert, always invert.Carl Jacobi, as Charlie Munger liked to repeat it