One of the central arguments in a recent edition of the Superhero.tech newsletter by Wojtek Strzałkowski and Piotr Kacała is that access to model intelligence is becoming increasingly commoditized. The observation is simple, but it changes how I think about where durable value in AI products actually comes from. It also matches what I have been watching from the product side for a while now - so I want to unpack it and add my own read.
The edge that keeps shrinking
The advantage of simply having access to a better model is already getting harder to defend.
Frontier models increasingly overlap in their capabilities. The gap between the best model and the second-best keeps narrowing, and for most practical product work the differences are no longer the deciding factor. What one lab ships this quarter, the others match within months - sometimes weeks.
Open-weight models continue to improve. Use cases that recently required much more expensive proprietary models can now be served by models anyone can download, fine-tune, and run on their own terms. The capability you once paid a premium for becomes something you can self-host.
And the leading AI labs are moving further into the application layer themselves. The companies that make the models are now also building the assistants, the agents, and the workflows on top of them - competing directly with the products that were built on their APIs.
The result is that access to a powerful model - or a polished wrapper around one - is a weaker moat every quarter. If your product's main ingredient is something everyone can buy, and the supplier is moving into your market, the ingredient is not the moat.
In construction, this difference is easy to see. Producing a progress estimate with AI is useful. Producing one that an owner and a contractor are both willing to rely on during commercial settlement is a different product problem entirely - and the second one takes much more than model intelligence.
The question worth asking instead
The more important question is: what part of the product cannot be copied or replaced overnight?
A clean interface can be rebuilt in a sprint. A clever prompt chain can be reverse-engineered from the outputs. A feature that demos well can be matched by a competitor watching the same demo. None of these are worthless - they matter for adoption and experience - but none of them survive contact with a determined competitor and a capable model.
The harder things to copy are the ones accumulated over time: domain knowledge, proprietary data, integrations, operating procedures and trust.
The value sits around the model
From my perspective, whether a company develops its own models or builds on top of existing ones, the defensible value increasingly sits around the model:
- Deep industry know-how. Knowing how the work actually gets done, including the exceptions that rarely make it into formal documentation. Learned in the field.
- High-quality, domain-specific datasets. Not just more data - historical cases with the labels, context and edge conditions specific to the industry, collected and cleaned over years.
- Integrations. Connections to the tools, file formats and processes an organization already runs on, often built through years of customer-specific work. Slow to build, slower to replace.
- Compliance with established procedures. In regulated environments, fitting the required process matters as much as model capability.
- Security, governance and standards such as ISO. Certifications take audits, documentation and organizational discipline to earn, and procurement departments increasingly treat them as entry requirements.
- Traceability and repeatability. A result needs to be reproducible months later, not only convincing when first generated.
- Commercial usability. The strongest outputs can support acceptance, settlement, payment or dispute resolution between parties.
A competitor with the same model and a better designer can copy your screens quickly. Reproducing a decade of domain data, certified processes, or a position inside a client's workflow is a different problem - and anything slower than overnight is what the current pace punishes.
What raw intelligence cannot provide: trust
Together, these things create something more valuable than model access: trust.
In many industries - especially those involving infrastructure, construction, finance, healthcare or regulated operations - an impressive answer is not the product. An AI output becomes valuable when the result can be verified, reproduced, audited and used as a reliable basis for a decision, an approval, a payment or a contractual discussion.
I see this daily in my own corner of the market. The intelligence generates the progress measurement. Everything around the intelligence is what makes both parties willing to use it.
Demos tend to hide this problem because they optimize for the successful case. A model that is right 95% of the time can be excellent in one product and unusable in another; the important question is what happens in the remaining 5%. If those errors can affect a payment, an approval or a contractual decision, the product needs a way to expose uncertainty, verify the result and reconstruct how it was produced.
From useful tools to operational infrastructure
For me, the useful distinction is not between AI products with better or worse models. It is between tools that assist a task - genuinely useful, easily adopted, and easily swapped for whichever alternative is cheaper or newer this quarter - and systems that organizations run their actual operations through, where the outputs feed decisions, approvals and money.
That is where AI products can move from being useful tools to becoming operational infrastructure. Once a product sits inside approvals, reporting or payment workflows, reliability matters more than how impressive the underlying model looks in a demo.
Model capability will probably keep getting cheaper and easier to access. Building enough trust for a customer to base real operations on the output is still expensive and slow. If I were placing product bets right now, I would place them on the product a client cannot rip out without rewriting their own procedures. That is the part a new model release does not immediately make obsolete.
Sources
- Superhero.tech newsletter - Wojtek Strzałkowski and Piotr Kacała, on the commoditization of access to model intelligence.