Drive the 101 freeway in San Francisco, Woodson Martin said, and the billboards for AI software companies all say "the exact same five words" with "just a different logo." Martin, chief executive of OutSystems, a platform that large organizations use to build business applications and, increasingly, AI agents, used that image to describe a problem facing his whole industry: everyone is starting to sell the same thing.
Martin made the argument on The Cognitive Revolution, in an episode published October 7, 2026, with host Nathan Labenz. It is the case one software incumbent makes for why established vendors, not AI-native startups, should come out ahead, and OutSystems was also one of the episode's sponsors.
Building blocks and the startup threat
Labenz opened the question with a pattern he said he has seen often. When AI arrived, the expectation was that startups built around it from day one, often called AI-native, might come in and win the market. Instead, he said, incumbents that already had "really well-developed primitives" usually found it easier to put an AI layer on top and get the most out of what they had.
Primitives here means the basic, reusable pieces of a software platform, such as how it handles data, user permissions and connections to other systems. An AI agent, a system that plans and carries out tasks using tools, can recombine those pieces quickly if they already exist and already work.
Martin agreed that those foundations matter and said primitives that work at enterprise scale are "really hard to build." But he said the software is only half of it. "The other thing is trust," he said.
Trust as a track record
Martin described the position of a regulated company operating somewhere in Europe: hundreds of regulators looking at its systems, contractual obligations to customers and business partners, and privacy concerns "everywhere you look." For a company like that, he said, trusting a system is "not just a pure technical evaluation of its underpinnings." It rests on the whole history of systems like the one it needs being delivered on that platform, and on being able to ask other customers what has worked and what has not.
That kind of reputation for "hardened systems in regulated industries," Martin said, "is a thing that is just very hard to get as a startup, regardless of how cool your technology is."
Everyone grows into the same product
Labenz then pushed on a second worry. Software companies, he said, feel "newly emboldened" to chase adjacent markets, and each one they enter opens up more. He sees platforms converging on the same "super horizontal, everything you need in one place" offering, and asked how anyone wins in that environment.
Martin called it the "sea of same." So many experiences can now be delivered through a conversation with an AI, he said, that the interesting question has moved to what sits behind the conversation: which systems, which data, and how they work together to give users what they want. Answering that quickly leads to the same shopping list, he said: an agent platform, an AI builder and integrations with many other systems. A lot of platforms, he said, are rushing to fill exactly that space.
Martin's answer was that "ultimately, it's going to be specialization." Everyone will need a common set of building blocks, he said; the difference lies in what a company gets really good at. He pointed to two kinds of specialization.
The first is in the AI models themselves: tuning and distillation, a technique for transferring a large model's abilities into a smaller one, along with model choices and cost structures matched to particular jobs.
The second is industry- and workflow-specific work. For OutSystems, Martin said, much of that lies in regulated industries where the company has already solved many problems. He described it as what he called 37 features that are not on the basic list but that a customer needs "to actually get through the hurdle of building or deploying the first thing."
Where OutSystems says it specializes
Martin listed the industries where he said OutSystems has built specialization over years: banking, insurance, government, healthcare, transportation, logistics, and energy and utilities. He called them all big regulated industries and the core of the company's current customer base, where its teams have specialized skills and can help organizations modernize processes quickly.
The company's own product page for industry solutions describes what that looks like as a product: a pre-built agentic system for a high-value process in a given industry, which arrives with the steps, the agents, connectivity to a customer's existing systems and an audit trail, and which the customer's team can adapt.
Martin did not claim the industry has solved its message problem. Right now, he said, vendors are all saying "we got all the stuff," and they are learning that this is "just confusing everybody." "We have work to do as an industry there," he said, "to get our game on."
Industry knowledge as the bottleneck for new hires
The same theme came up when Labenz asked about hiring. Martin said he is "super bullish" on junior employees who have spent recent years native to AI tools and bring an "AI-pilled" approach. But he said organizations have not yet built the infrastructure to teach them the industry processes and specialized knowledge they need to put those skills to full use.
OutSystems, he said, is leaning into the model of forward-deployed engineers, who work directly with customers to imagine and design the work. An OutSystems practitioner, Stefan Weber, has described that role as engineers embedded with business stakeholders who share responsibility for working out requirements and delivering results. Martin said the company is also shifting new-hire onboarding, long a few days of orientation plus follow-up training, toward a more agent-driven, just-in-time approach that gives employees knowledge when they can use it.