AI Strategy
The Intelligence Layer: Why the Next SaaS Moat Is Context, Not Features
Key Takeaways
- •Nearly nine in ten organizations now use AI in at least one business function, making AI capabilities table stakes rather than differentiators.
- •Only 5% of companies capture substantial value from AI while 60% generate almost none, according to BCG research on AI value generation.
- •Features built on foundation models can be replicated by competitors in as little as eleven days or a single weekend because they use the same underlying models.
- •The intelligence layer comprises accumulated customer history, pipeline data, campaign results, decision rationale, and buyer behavior patterns that competitors cannot download or replicate.
- •Context-rich systems create data flywheels where every interaction generates proprietary data that feeds back to sharpen subsequent AI outputs and decisions.
- •McKinsey identifies privileged data as a competitive moat because models reasoning over unique datasets produce outputs competitors cannot match with identical models.
- •Intelligence layers compound in value over time as they accumulate more structured context, while static features depreciate immediately after launch as competitors catch up.
- •AI-native companies that centralize intelligence in shared layers rather than siloed tools belong to the 5% capturing substantial AI value, according to organizational AI research.
A founder I know shipped a slick AI feature last quarter. He was proud of it, and he should have been. His team scoped it on a Monday, built it with a model and a weekend, and had it live by Friday. Customers liked it. He told me it was going to be his differentiator.
Eleven days later, a competitor shipped the same feature.
Not a similar one. The same one, down to the interaction pattern, because they were both standing on the same foundation models and the same playbook. The thing he thought was his edge turned out to be a thing anyone could rent for a few dollars per thousand tokens. He had built a feature. He had not built a moat.
This is the quiet crisis underneath the AI boom, and it is not really about features at all. McKinsey's latest work on competitive advantage opens with a line borrowed from a movie: if everyone is special, then no one is. Nearly nine in ten organizations now use AI in at least one business function, and most of them are deploying the same large language models to do it. When everyone has the same advantage, it stops being an advantage. The proof is in the value gap. BCG found that only 5% of companies are capturing substantial value from AI, while 60% are generating almost none. The model is not the divider. Something underneath it is.
Features Are Now the Cheapest Thing You Own
For thirty years, software companies competed on features. You shipped something the other guy did not have, you charged for it, and you defended the gap with engineering time. The whole category was built on the assumption that capability was expensive to copy.
AI broke that assumption. When a competitor can stand up a comparable feature in a weekend using the same models you used, the feature is no longer a defensible asset. It is a commodity with a short shelf life. McKinsey puts it plainly: apps and tools can be copied, and value comes from building advantages that are hard for competitors to replicate. The same thing happened during the web and mobile waves. Everyone rushed to build websites and apps, and the leaders did not pull ahead just by having them. They pulled ahead by weaving the new capability into something deeper that the laggards could not duplicate.
So if the feature is not the moat, what is. The answer is the thing the feature runs on top of.
The Durable Layer Is Context
What a competitor cannot copy over a weekend is everything your business knows about itself.
Your customer history. The shape of your pipeline. The campaigns that worked and the ones that quietly failed. The reasoning behind decisions you made two years ago. Your voice, your positioning, the specific texture of how your buyers actually behave. None of that is downloadable. It is accumulated, structured, and yours, and it is the only thing in the AI stack that gets more valuable the longer you operate. McKinsey calls privileged data a moat for exactly this reason: when a model reasons over data competitors do not have, every interaction generates more of that data, which feeds back and sharpens the next answer. They call it a data flywheel. I call it the intelligence layer, and it is the layer that matters.
I have made a version of this argument before about output quality. In Context Is the Whole Game I argued that the model is commoditizing and the context wrapped around it is where the value actually lives. The moat argument is the same insight pointed at strategy instead of content. A thin-context company and a rich-context company can buy the identical model. One produces fast generic work that anyone could produce. The other produces work that is specific to its business in a way no competitor can reach, because the competitor does not have the context to reason from. The model is the same. The intelligence layer is not.
Why the Intelligence Layer Compounds
The reason context is a moat and a feature is not comes down to one word: compounding.
A feature is static. The day you ship it, it is as good as it will ever be, and it starts losing its edge immediately as others catch up. An intelligence layer moves the other direction. Every campaign you run, every decision you log, every customer interaction you capture makes it richer, which makes the next decision sharper, which produces a better outcome you can learn from. This is the difference between a system that answers and a system that learns. I drew that line in RAG Is Only Half the Story: retrieval lets AI look something up, but living intelligence accumulates, so the asset is worth more every quarter rather than depreciating like code.
This is also why the moat is structural, not technical. In The AI-Native Company I argued that the companies pulling ahead are the ones whose intelligence flows from a shared layer instead of living trapped in scattered tools and individual heads. That is the intelligence layer described from the inside. The 5% capturing real value did the unglamorous work of making their context usable and connected. The 60% bolted clever features onto a structure that forgets everything the moment a task ends, and they are now watching their differentiation evaporate every time a competitor ships the same weekend build.
What to Build This Year
If features are commodities and context is the moat, the strategic priority changes.
Stop measuring your roadmap only by what it ships and start measuring what it accumulates. Ask a sharper question about every initiative: when this runs, does it make our intelligence layer richer, or does it just produce a one-off output and forget. A chatbot that answers and forgets adds nothing to the moat. A system that captures the interaction, structures it, and feeds it back into a shared context that every future decision can reach is building the asset. As I put it in What Is Operational Intelligence, the point was never to move faster. It was to make the organization's knowledge legible and reusable, so intelligence sits upstream of the work instead of being rebuilt from scratch each time.
The companies that win the next few years will not be the ones with the cleverest feature, because the cleverest feature has a two-week half-life. As McKinsey concludes, advantage will come from the organization that turns common models into uncommon moats faster than anyone else. The uncommon part is never the model. It is the context only you have.
The Takeaway
Features stopped being a moat the moment your competitor could rent the same model you did. What they cannot rent is your accumulated, structured context: who your customers are, what has worked, how you think, why you decided what you decided.
That intelligence layer is the one part of the AI stack that compounds instead of depreciating. It gets richer every time it is used, and it is the thing no weekend build can copy. The model is table stakes. The features are commodities. The moat is context, and the companies that treat it as their core asset rather than an afterthought are the ones that will still be differentiated when the current feature race is over.
In the AI era, you do not own your software. You own what your software knows.
Tracy Thayne is the founder of Expona, an AI consulting practice for smaller B2B companies in construction, manufacturing and the trades around them. Take the free Decision Line Assessment or subscribe to the blog (below) for plain spoken posts on putting AI to work in businesses that don't call themselves tech.
Frequently asked questions
Why are AI features no longer a competitive moat for SaaS companies?
AI features are no longer moats because competitors can replicate them in days or weeks using the same foundation models. When everyone accesses identical large language models, features become commodities with short shelf lives rather than defensible assets. The model itself is table stakes, not a differentiator.
What is the intelligence layer in AI-powered software?
The intelligence layer is a company's accumulated, structured context including customer history, pipeline shape, campaign performance data, decision rationale, positioning, and buyer behavior patterns. This proprietary operational knowledge cannot be downloaded or rented by competitors and becomes more valuable over time as it compounds through continued use.
How does a data flywheel create competitive advantage in AI systems?
A data flywheel creates advantage when AI models reason over proprietary data competitors lack, and each interaction generates additional unique data that feeds back to improve subsequent outputs. This compounding effect makes context-rich systems progressively more valuable, while competitors using the same models without proprietary context produce only generic results.
What separates the 5% of companies capturing AI value from the 60% generating almost none?
The 5% capturing substantial AI value built intelligence layers that accumulate and structure context, making organizational knowledge reusable across decisions. The 60% generating minimal value bolted features onto systems that forget after each task, failing to build compounding assets that differentiate their AI outputs from competitors using identical models.
What should SaaS companies prioritize instead of feature development in the AI era?
SaaS companies should prioritize building systems that accumulate and structure context rather than shipping one-off features. Every initiative should be evaluated on whether it enriches the intelligence layer and feeds proprietary data back into shared organizational knowledge, creating compounding assets competitors cannot replicate by renting the same models.
Subscribe
Get notified by email when we publish a new post. No spam, unsubscribe anytime.