Running a business with AI has a meter that's always running. Every question, every text and every bit of code ticks through in tokens. What that meter doesn't count is meeting people.
Running a business with AI: what does the meter actually count?
A while ago I had everything calculated. In June 2026 I consumed 564 million tokens in Claude Code. In September it was 5.76 billion, spread across 21,826 prompts. Ten times as much in three months.
What do I get in return? Speed and focus, certainly. But increasingly also texts that don't make me happy at all. Tidy, neat and precisely according to the norm of the average. Logical, because a language model is trained on texts from all over the world, some of which have simply been copied. It gives you the average of everything. Researchers from UCL and the University of Exeter saw it happening already in 2024. Stories written with the help of AI began to resemble each other more and more.
This phenomenon is AI slop. Instagram is full of this hashtag. "This creepy, bizarre and demonstrably fake content is often called 'slop'", according to Merriam-Webster. It defines it as cheap content in large quantities that's 100 per cent made with AI. So that's the rubbish now being prompted into designs. And what about the linguistic impoverishment right now? Worldwide, all texts are beginning to resemble each other more and more. Who's going to read that slop in six months' time?
Jim Rohn once said: "You are the average of the five people you spend the most time with". If one of those five is a language model, you automatically become the average of everything. Then it certainly helps to see those other four in person a bit more often.
Last Wednesday I was cycling to work and I bumped into one of the people I work with a great deal. He was also on his way to his office. We get off and stand chatting for ten minutes. It was a nice conversation. When I get back on, I have ideas for two existing clients, about branding and marketing together. And immediately a brilliant idea to go for lunch and tackle a few things in my communication.
That's the contrast with the grind you experience when you try to solve everything with AI. Prompt, answer, another prompt. The result is fine, but it feels like a dopamine hit: a quick kick, and then the next one. That's not down to me, that's how a human brain works. It hunts for the next reward, and the machine always delivers it, day and night. So I stack up billions of tokens and wonder whether I'm still working or scoring. A meeting with a human being is worth a thousand times more, and it doesn't appear on any meter. It's precisely those meetings that give life value. When running a business with AI you forget that quickly, because the screen always gives an answer.
A person in the street gives a different answer. Spontaneous conversations are good for you, we've known that for centuries. As far back as 1847 there was a man walking through Copenhagen every day who talked with whoever he met. He called it a people bath. That man was Søren Kierkegaard, the Danish philosopher. With me it too often doesn't happen, because I'm in the AI machine myself, just like thousands of others. Yet this is the division that works. Running a business with AI only delivers something when the machine does the executive work and you collect the ideas from people. The judgement remains yours, as I wrote earlier about what no algorithm will ever do for you.
In September the meter counted 5.76 billion tokens. Ten minutes beside the bike cost nothing and were worth more.
Sources
Anil R. Doshi and Oliver P. Hauser, "Generative AI enhances individual creativity but reduces the collective diversity of novel content", Science Advances, 2024.
"Merriam-Webster's 2025 word of the year is 'slop'", Associated Press, 15 December 2025.
Jim Rohn, American entrepreneur and speaker (1930-2009). The statement about the five people is widely attributed to him.
Søren Kierkegaard was a Danish philosopher (1813-1855). He wrote about his daily walks and conversations through Copenhagen in his diaries. And in a letter to Henriette Lund from 1847.