When people talk about using AI in business, the answer usually sounds simple.
AI writes code. Generates images and videos. Helps designers, marketers, analysts.
In my own team, AI is already part of everyday work. We partially write code with it. Use it in support. Inside internal knowledge bases. For data enrichment. And for dozens of small, boring tasks that used to eat time.
In practice, AI today mostly lives in the digital world.
It helps digital businesses move faster, cheaper, and more efficiently.
All of this looks impressive. But lately I keep catching myself on the same thought: maybe we are still at a very early stage.
Because almost everything we currently call "AI adoption" happens in a safe environment.
Texts. Interfaces. Documents. Code.
Places where mistakes are relatively cheap.
Recently, I have been noticing more projects that are interesting not as another SaaS tool, but as attempts to embed AI into real processes: production, materials, supply chains, regulations, physical constraints.
For example, Keychain is a platform that helps retailers and brands launch private label products. AI connects a product idea with real factories: ingredients, production capabilities, requirements, volumes.
This is no longer about generating a pitch deck. This is about working with actual manufacturing.
Or Drafted: a platform where a person can design their future home with the help of an AI agent. Not just a visualization, but a path from an abstract idea to a project that could, at least theoretically, be built.
One platform that made a particularly strong impression on me is One.five.
AI there is not a single assistant. It is a group of agents, each playing a professional role: materials engineer, supplier and pricing analyst, packaging designer, regulatory expert for different countries.
In practice, it looks like a simulation of a full specialist team, assembled inside one digital space.
In these examples, AI is no longer just helping a human.
It becomes a coordinator of the real world. A layer that connects ideas, people, companies, documents, materials, and physical processes.
This is a fundamentally different level.
There is money here. Contracts. Deadlines. Responsibility. Mistakes that are expensive.
AI starts working not where experimentation is cheap, but where every decision has weight.
And it feels like the potential here is almost unlimited.
Packaging for any kind of product. Furniture manufacturing optimization. Construction, with all its bureaucracy, documentation, contractors, and supply chains. Product lifecycle management from idea to disposal.
Everything that today relies on human coordination, spreadsheets, emails, and endless calls.
Not as a replacement for humans, but as a connective layer between chaotic human thinking and the rigid logic of the physical world.
My bet is that in the coming years we will see many startups exactly in this space.
Not another chat or assistant, but AI as an operator of reality. One that proposes options, checks constraints, matches data, and removes a massive amount of manual, exhausting work.
And yes: I want to be part of this process too.
It feels like the most exciting part is only just beginning.