Every founder needs an AI strategy (it can be one page)

Every week, someone sends me a massive presentation about their company's new AI direction. It usually involves a lot of boxes, arrows, and terms that did not exist six months ago. I read them because of the work I am doing with Substrat, helping builders set up their AI infrastructure. But I will be honest with you. Most of them are a complete waste of time. They read like someone is trying to convince themselves they are safe.

When Fredrik Bage and I were building Mediapilot back in Stockholm, we used to get bogged down in long-term planning. I had met him years earlier when I was grinding out commission-only sales at a PR agency against everyone's advice, and he was the one who told me I would be good at it. We grew Mediapilot over roughly seven years and created more than a hundred jobs, but our best decisions were rarely the result of a fifty-page slide deck. They came from looking at an immediate, obvious shift in how our clients behaved, and adapting to it before we ran out of cash. Later, when I worked on the commercial side at Happeo, it was the exact same story. The market changes, and you react. AI feels bigger, but the mechanics of adapting are the same. You do not need a committee or an external agency to build an AI strategy. You need a single piece of paper.

Three questions for a single piece of paper

I have a habit of keeping a written life plan. I started doing it around 2010, rating my health, finances, work, personal time, and relationships out of ten every year. It forces me to confront reality on one page. A good business strategy should do the same. If you are running a company right now, your entire AI strategy should consist of your honest answers to three questions.

First, what does this make free? Think about your core operations. In my early twenties, after coming back to Sweden from studying in the United States, generating targeted sales outreach took hours of manual work. Today, the marginal cost of generating text, code, and basic analysis has essentially dropped to zero. You have to assume that any process in your company that relies on summarising information or writing boilerplate code will soon cost nothing. If your business model relies on charging clients for those specific tasks, you have a massive problem. Write down exactly which parts of your daily operations are about to become free commodities.

If the cost of creating a standard product drops to zero, the value immediately shifts to the person deciding what to build in the first place.

Second, what does this make scarce? This is the harder question. When something becomes abundant, its opposite becomes valuable. If perfect grammar and endless content are free, then human taste, trust, and physical presence become incredibly scarce. I am currently building Anchor, an AI tool for alienated fathers. We use AI to handle the legal logic and the heavy lifting, but the actual value we provide is trust and emotional clarity in a very dark moment. If the cost of creating a standard product drops to zero, the value immediately shifts to the person deciding what to build in the first place, or the person checking if it is actually any good.

Third, what would a competitor built today look like? Imagine someone starts a company in a garage tomorrow. They have no legacy software, no office lease, and no established way of doing things. They just have current open-source models and a strong desire to take your customers. How do they build their machine? What do they automate from day one that you are currently paying twenty people to do? Write it down. That imaginary competitor is exactly who you are fighting.

Expiry dates and honest reviews

You cannot write this document and put it in a drawer. The underlying technology is shifting too quickly. We moved our family down to Marbella in 2018 meaning to stay for two years, and the only reason we are still here is that we kept re-evaluating our setup. Every few months, we looked at our situation and decided it still worked. You have to treat your one-page strategy with the same ruthless frequency. Review it every single quarter. Pour a coffee, or maybe a glass of wine, take out the paper, and read it.

Ask yourself if your answers hold up against what actually happened over the last ninety days. Usually, you will find that a capability you thought was two years away is already running on your phone, while something you assumed would be solved instantly is still completely broken. I haven't got this figured out. I play a lot of tennis down here, usually quite badly, and the AI space feels exactly like playing against someone much better than you. The ball comes back faster than you expect, at an angle you did not prepare for. I regularly look at my own assumptions for Substrat and realise I was entirely wrong about how builders want to interact with a new framework. But writing it down in plain English, on one page, means I can actually see the mistake and fix it. That is all a strategy is anyway. A best guess, written down clearly, waiting to be corrected.