I run an AI agency from Dubai, and across the whole business I keep 10 to 15 skills in regular use. Most of the people whose setups I see have built far more than that, and they built them too early. A bare chat gets me to 80 percent in a minute, and no skill was involved in getting there.
I saw the cost of the other approach on a consulting call with the owner of a law firm. He had built a skill for almost every recurring task, and that collection had quietly turned into his main complaint. He no longer knew what he owned, so every piece of work now began with a question about which of his own tools applied. The way he put it himself was that he had no structure.
The reflex that produces too many skills
The path there makes complete sense, and I suspect almost everyone walks it. You are working, you hit a bottleneck, and you fix it immediately by turning it into a skill. Next time you will not start from zero, which feels like an obvious win in the moment.
That is exactly why the reflex is so hard to spot. Every single decision was the right one, and the outcome is still wrong, because the decisions accumulate. Do it for enough separate topics and you have lost the overview of your own collection.
So the effort does not disappear. It changes position. You win a minute while working and you pay it back while looking, and you pay it back every time.
What a missing structure actually costs
The lawyer described it better than I could. What bothers him in daily use is that he no longer knows what he already has. He then has to think far too much about it before starting.
His own example was a shareholder agreement. Before he can start on one, he has to work out which skill he built for it, what that skill covers and where it stops. The same question comes back for his finances, and it comes back again for every further topic.
That is the real invoice, and it appears on no dashboard. Every one of those decisions spends attention before the work itself begins. A tool you have to remember first is more expensive than no tool at all.
Why a bare chat already starts at 80 percent
My basic way of working is that a bare chat gives me good results inside a minute. I never start at zero. I start at 80 percent, and that has nothing to do with clever wording or with anything I built in advance.
It works because nothing of mine lives in a chat. I work on one task, save out whatever I need, and then delete the chat again. In the evening I close all of them, and I still lose no knowledge at all.
My knowledge sits in folders, which is why every new chat can find it. Leave your knowledge inside chats and artefacts instead, and getting it back out later becomes close to impossible. A lot of people are going to lose a lot of knowledge that way without noticing.
When a skill is genuinely worth building
For me to build a skill, more has to come together than a single bottleneck. It has to be a task that comes back regularly and that I handle the same way every time. Everything else goes to a plain chat, and the plain chat handles it well enough.
Behind that sits a division of labour I explain on every client call. The AI is the construction worker and you are the architect. Plenty of successful people now use AI for everything and let it make decisions as well, and that is the part where I stay careful.
An architect who builds a new tool for every wall never finishes the house. Building the tools feels like progress because the tools are visible, and that is the trap in it.
Structure beats tooling
Ten to 15 skills carry an entire business, provided the knowledge sits in folders rather than in chats. Building more usually means replacing a missing structure with additional tooling, which moves the work instead of doing it.
The question underneath comes up in every client conversation I have. What happens when today's model is no longer the best one tomorrow? Knowledge that sits in folders survives that switch, and knowledge that only sits in chats does not.
So the next step is unglamorous, and it has nothing to do with tooling. Move your knowledge out of your chats and into a folder structure before you create the next skill. After that you can see far more clearly which skill is still worth building at all.