Build business systems with AI sounds like a headline until you watch how fast it moves. On day four of my 30 day experiment, the speed was the first thing I noticed.

Today is consecutive day 2,146 of my running streak. I have now covered 16,484.28km towards a lap of the world, 40,075km in total, which leaves 23,590.72km. I run every day in barefoot-style footwear, and I film an episode every day. The aim is to raise £1M for children's causes. That daily rhythm is the backdrop for everything else I do, including this experiment.

What the experiment is

I am testing whether I can take a business concept from build, to launch, to delivery, to profit in 30 days, with AI doing around 90% of the workload. My part is limited. I connect AI to tools it cannot reach by itself, I set certain settings, and I give feedback so it can run as autonomously as possible.

The concept comes from my own background. I used to coach and consult with micro and small business owners on systems and processes. I was going to offer bespoke operating systems for businesses. Then AI changed the maths. Something that took me 18 months to build, I believe I could now replicate in a week or two, if I know what I am doing with AI.

So the idea has changed. I want to build a marketplace of business operating systems, built with AI under my direction and installed into a small business owner's own setup. I am leaning towards one-off fees rather than subscriptions, so the owner holds the system once it is installed. That is the plan, and it is still a plan. Nothing has been sold yet, and I will not pretend otherwise.

What happened by day four

Here is what exists after four days:

- A real-time profit and loss system. It takes live bank transactions, reconciles them, categorises them the way the owner wants, and updates the profit and loss as it goes. It has graphs and visual dashboards on top. I built the first version in around half a day. - A task and project management system. It tracks progress against projects and goals, lets you delegate and comment, and uses AI to put work in priority order based on your capacity and your calendar. I reached a first version in around half a day. I expect another half day of tweaks. - A leadership dashboard I had already started. It shows the key metrics for a business, adds AI commentary and recommendations, and runs a 31 day cash flow forecast with best case, worst case and actual views. - An inbound communications system. It sorts an email inbox, drafts replies, and picks up invoices so they feed an invoice dashboard. It is similar in spirit to tools like Fyxer, and I have pushed it further by linking it to bank transactions.

My team has also set up a database to test how we onboard clients, plus a WordPress and WooCommerce setup connected to Stripe and AI, ready for the new website over the next few days.

Nothing took more than a full day. Most pieces took hours.

The lesson for me

The first lesson is about speed. My honest estimate is that AI is making this at least ten times faster than doing it manually. That is my own estimate from experience, not a measured study. I would rather say that plainly than dress it up.

The second lesson is about what stays human. The speed does not remove the need for feedback and iteration. AI builds quickly, but it builds what I direct. Years of working on systems and processes mean I know what a good profit and loss view should show, and what a task system needs to do for a team. The experience still matters. It is applied differently.

The third lesson is about perspective. I am ahead of schedule at day four, but day four is early. The hard parts are launch, delivery and profit. A fast build tells me nothing about whether anyone wants to buy it. I will judge the experiment on the later days, not this one.

There is a question I asked myself during the episode. Can I afford not to use a tenfold improvement in productivity? For me, the answer is no. I would add that this applies to anyone running a small business who is stretched for time.

How it connects to the streak

The running streak taught me that consistency compounds. One run means little. Two thousand one hundred and forty-six consecutive days is a different thing. The AI experiment follows the same logic. One day of building proves little. Thirty days of steady, directed work will tell me something real.

The mission sits behind both. I run and film every day to put the fundraiser in front of more people, because more people watching means more money raised for children's causes. The £1M target is the reason I keep going on days when I would rather not. Any business that comes from this experiment is separate from that mission, but the discipline I use for one feeds the other.

Where this leaves me

Day four is a good day, and I am treating it as a single data point. I will keep recording what works, what does not, and what the numbers say as the 30 days go on.

If you run a micro or small business and you are curious how far AI can take your systems, follow along. If you believe in the mission, please subscribe, share, like or comment on the episodes. Every extra person who sees them helps the fundraiser for children's causes.

I will be back tomorrow with day 2,147.

Watch the full episode: https://youtu.be/j6B3I7HG7qg