Today is day 2,140 of my daily run. I have covered 16,440.93km so far, which leaves 23,634.07km of the 40,075km lap of the world. I run barefoot style every day to raise £1M for children's causes. The streak is planned over 16.5 years, so I judge my decisions by whether I would still stand behind them years from now.

Today's question is one many founders ask. Is it possible to start a business using only AI?

I know the answer in general. Plenty of people already do this. The open question is whether I am able to do the same with what I know. So I am running an experiment, and I will report what happens, including the misses.

Where I stand with AI

I am not at the front of the field. I would call myself an early adopter, somewhere between average and advanced. I have used AI tools since ChatGPT first launched, and I have worked across several platforms since. That gives me working capability, not expertise.

What I decided against, and why

For 18 months I built Operations Director, a system for micro and small business owners covering real-time financial forecasting, systemisation, processes and team management. Version 1 went through beta testing. I took the feedback, adapted the product, and prepared a version 2 launch.

Then I chose not to launch. I had two reasons.

- I wanted to focus on my property business. - AI was changing software businesses, consultancies and agencies. Services once sold at high-ticket prices were becoming available at a fraction of the cost.

I also saw that what took me 18 months might be rebuilt in a week or two with the latest models. Holding a launch back felt like the right call at the time.

What changed

Several months later, I see many people bringing AI-built products to market and doing well. I already have a concept proven and a business mapped out. The product is not the gap. The market conditions are different, so my approach needs to be different.

So I am going back to Operations Director and launching in a new way.

How the model changes

The original plan was a software consultancy. We would build each client's operating system and then provide ongoing support. That model is labour intensive and carries high-ticket pricing.

The new plan is this:

- Break the platform into standalone modules, each solving one problem for a micro or small business owner. - Charge a one-off fee per module. - Sell support separately, in chunks, for owners who want it. - Move away from subscriptions, because retention is hard in the current software market.

I believe a one-off fee suits a small business better than a fixed monthly cost. I have not tested this yet, so I treat it as a belief.

Who does the work

Two members of my team will help me. They will code and build using AI, so three people will be using AI to create this business. My target is to launch within one to two weeks. I know this is a bold target. I also still run a property business. The heavy part of that work is ending and the focus is moving to maintenance, which is why the timing works. I am conscious of spreading myself too thin.

The aim is for AI to do the building, the delivery and the support. A client should be able to buy through the website and be onboarded without human intervention. Marketing campaigns are in scope too. I will give AI the supporting material I have and ask it to produce the business plan and the launch steps. Even the process of turning an idea into an action plan becomes a product. The result is a supermarket of AI-built solutions.

The reason to work this way is scale. The consultancy model capped how many clients we served, because each one took manpower. Modules built and delivered by AI remove this cap.

Why the market makes sense to me

My understanding is that over 75% of micro and small businesses are one-person operations. Those owners lack the time and headspace to learn AI themselves. When we spend our days in this world, we assume everyone else does too. Most people do not. An off-the-shelf solution, set up and explained for them, meets a real need.

The lesson for me

Two things stand out.

First, a decision made on the facts of one month is allowed to change when the facts change. I said no to launching, and now I am saying yes. I would rather say so openly than defend an old position.

Second, the run keeps me honest. A streak of 2,140 days is built from one run at a time. A one to two week launch target is only worth something if I measure what is done each day, not what I hoped to do. Consistency over intensity applies to business as much as to running.

Why this connects to the mission

Each day I run, record and share is part of the fundraiser for children's causes. The more people this reaches, the more support the mission has a chance to gain. Running a business well, while keeping the streak and my family time protected, is part of making the full 40,075km possible.

Next week I will keep sharing the process, including what works and what does not.

If you are starting a business or running one, tell me what you would want an AI-built business to handle first. Watch the full episode, follow the journey, and if you believe in the mission, subscribe so more people see it.

Watch the full episode: https://youtu.be/BB30e4u0F4o