My bet for 2027: fewer new startups building B2B applications that support a company's full workflow. More companies building their own workflows, on top of infrastructure they still happily pay someone else to run.
A company buys SaaS because it needs a process to work. As building gets easier and cheaper, I expect more companies to make the workflow fit their business themselves. (There are already some early signs supporting this in McKinsey's 2026 AI survey.) Payment infrastructure, authentication, secure storage and hosting are still things they'll want to hand off. I expect more of the tech startup space, and probably more VC money, to shift toward infrastructure and lower-level components.
A year ago, we were wondering whether AI would kill the service industry. I now think this shift could create more work for providers who know how to build with AI. Building in-house can still involve a lot of outside help: someone has to connect the systems, get the workflow into production and support it afterwards.
Enterprises already turn to service providers to understand what's possible and which direction to take. I expect the current pace of development to widen that gap. Put that together with more workflows built in-house, and I think AI-native services will grow strongly in 2027. The useful provider will need to help choose a direction and stay involved long enough to get it working.
I also think we could see higher valuations for providers that combine the right AI-native skills with software that stays useful at the customer. Utilities, frameworks and deployment tools can remain in use long after the first project. A growing installed base gives the provider a way to keep earning through subscriptions, maintenance and support, while improving tools it can reuse across customers.
Wonderful stands out to me as a particularly successful example of this model in Europe: embedded deployment teams working with a shared AI platform.
