The value of AI is not more content. It is less fog.
I am not interested in AI because it writes a better product description.
Good if it does. Nice.
But it is not a big thing.
In a product business the real value of AI is not more content. We have more than enough of that already. More reports, more dashboards, more advertising data, more product listings, more channels, more Excel files and more separate truths than anyone has time to actually take in.
The internet has the same problem, measurably. Graphite tracked article publishing and found that since November 2024 machines publish more articles than people do. The same study found most of those articles never appear in Google results or in ChatGPT answers. Content nobody asked for, read by nobody. So no, I do not need AI to make more of that.
The problem in a product business is not lack of data anyway.
The problem is that money gets committed to decisions before the financial consequence of the decision is properly visible.
This is the point where AI starts to be actually interesting.
This year our roughly £3 million product business runs on three people and an AI operating system. Not because AI is trendy. Because I want the consequence visible before the money moves.
Most of the hard decisions in a physical product business are not very romantic in the end. Buy more of this product or not. Increase advertising or not. Clear old stock at a discount or wait. Launch a new product or fix an old one. Believe last month's sales spike, or assume it was the campaign talking.
These decisions sound mundane, but the company's money gets tied to them.
If sales grow 20%, when do I need to buy more? Sooner than the bank balance suggests. If I increase advertising, will the stock last? If the stock lasts, will the cash last to the next purchase order? And if it lasts to the purchase order, does it also last to the freight, the duty, the salaries and the marketing? Which product looks good in revenue but quietly eats money?
There has traditionally been no single view to these questions.
The information has lived in different places. Sales in one system, stock in a second, purchasing in a third, advertising in a fourth. Cash on the bank account. Margin in some Excel. And in the founder's head the final logic that connects them all, available in the evenings and unreliable under stress.
I have lived in that world a long time. I have looked at a product and known it sells, and at the same time wondered if it eats too much capital. I have looked at a campaign bringing revenue and wondered whether that revenue actually makes sense. I have stood in a warehouse knowing there is money in the racking that could be doing something better somewhere else.
This is where AI can help.
Not by deciding for me.
By showing more clearly what a decision means, before the money is committed.
And most of the industry is not using it for this. IHL's research found fewer than one in four retailers have implemented AI anywhere near the places where stock goes wrong. The rest are, I assume, using it to write things.
A good AI operating system for a product company is not a chatbot. It is not a friendly interface where you ask how the business is doing and get a pleasant summary and a compliment back.
A good system helps you see where the next pound is worth tying. Which product produces in relation to the capital sitting in it, and which one only looks good because the revenue is high. Where the stock runs out if the advertising actually works. Where there is too much stock and the decision should have been made a month ago. I know that last one well, because I have been the person making it a month late.
This is lot more interesting than AI content.
Because the problem of a product company is not that nobody can write more text. The problem is that a wrong decision turns very quickly into goods, cost, hurry and stress, and by the time it is visible in the monthly numbers it has been true for weeks.
So what do I actually want from AI?
Fewer surprises.
The reality of the business one product at a time. The margin, the stock, the tied-up capital, the buying need, the advertising and the cash effect. Weekly. Early enough to act, because in a physical product business, too late is expensive.
The value of AI is not more content.
It is less fog.
And sometimes the most valuable thing AI can say is not "buy more".
It is: do not tie any more money into this.
Share of AI-generated articles and their visibility in Google and ChatGPT: Graphite, analysis of 43,000 URLs from Common Crawl, published 2025. Retail AI adoption: IHL Group, September 2025, fewer than 25% of retailers with AI or machine learning implemented in the areas most affected by inventory distortion.