Several companies, one pattern, and why I'm optimistic about AI

I founded CardioMood. I co-founded bilz.ai and iproxy.online. I work with Corsano Health as a technology partner, and I have started several telecom projects in virtual numbers and messaging. Written as a list, that looks like bets scattered across health technology, restaurant software, mobile infrastructure and telecom.

The companies share one shape. Each takes raw signals that people used to read by hand and turns them into something a person can act on.

The same shape, four times

• CardioMood reads a medical-grade bracelet worn 24/7 and alerts an executive coach when a client’s stress or recovery moves outside its normal range.
• Corsano Health makes medical-grade wearables that monitor patients continuously on general wards, at home after discharge and in clinical trials.
• bilz.ai reads supplier invoices line by line from a photo, a PDF or a forwarded email, and shows a restaurant what each dish costs and what the day earned.
• iproxy.online turns an Android phone and a SIM card into a private mobile proxy that its owner controls through a REST API or a Telegram bot.

The inputs are ordinary things: a bracelet, an invoice, a phone. The hard part is making the output reliable enough that someone runs a coaching practice, a hospital ward, a kitchen or a proxy business on it.

That shared shape is what makes several companies workable for me. The same questions come up in each one. How do you collect data from devices you don’t control? What happens when a phone drops off the network or a bracelet loses its connection for a week? How do you turn a noisy signal into an alert that a busy person will act on? An answer worked out for one company is a head start for the next.

Why I’m optimistic about AI

I believe one person working with AI agents can build a unicorn, a company valued at a billion dollars.

My reasoning starts with how software companies have grown so far. More customers meant more support staff. More documents meant more people typing them in. More data meant more analysts, and more people meant more managers to coordinate them. Headcount set the ceiling on what a company could take on, and raising money was mostly a way to raise that ceiling.

AI agents take on exactly the work that used to set that ceiling. I see it in the products I build.

At bilz.ai, reading every line of every supplier invoice was data entry for a person, and a busy kitchen has no one with time for it. Now the software reads the invoice from a photo and updates the ingredient prices. An owner can ask about the numbers in chat or by voice, or drop an Excel file, a PDF or a Word document into the conversation.

At CardioMood, correlation analysis runs across every client’s data around the clock, and each report comes with a written summary in one click. A coach would otherwise need an analyst to do that work, or would skip it.

An agent that reads invoices, watches data, drafts reports and answers common questions does work that used to need a department. Once agents also handle a large share of support, operations and engineering, a company’s size depends on the judgement of the people directing the agents: which problem to pick, what to build, what to leave out and when to change course.

I hold this as a belief, with no date attached. I think the pieces are arriving fast enough that it is worth building as if it will happen.

What that belief means for a new company

If one person with agents can go that far, a new company should be designed so that software does the routine work from its first version. The software reads the documents, watches the data, writes the first draft of the report and answers the question that comes up every day. People spend their time on the decisions that need judgement.

bilz.ai and CardioMood already work this way in parts of the product. Invoices are read by software, and reports are summarised by software. The coach decides how to intervene, and the restaurant owner decides what to do about a supplier who keeps raising prices.