Hey friend,
Growing up in Conakry, the power cut out most days. The surges and outages killed appliances constantly, and replacing them cost money we did not have. So you learned to open the thing up, find the one part that had actually failed, and fix that part. Not the whole machine. The broken piece. New things were expensive. Diagnosis was free.
I think about that every time I watch someone buy another AI tool.
A leader wires a shiny new model into the process they already run, and waits. Six months later the team is frustrated, the number has not moved, and everyone blames the model.
The model is usually fine. The map is wrong.
Every operation is a chain. Work flows from one step to the next, and somewhere in that chain one link does all the waiting: a handoff where work piles up, a step where the customer quietly gives up. Eli Goldratt called that link the constraint, and he showed something most people never quite believe. The constraint is the only place where improvement moves the whole system. Fix anywhere else and you are polishing a link that was never the problem.
AI does not change that math. It amplifies whatever you point it at.
Which means a wrongly aimed AI is not neutral. It is worse than what you had before, because now the stuck system runs faster and looks productive.
Point it at the constraint and you get real speed. Point it at a link that was already fine and you get a faster version of the same stuck system. And if the process was broken before AI ever arrived, you get confident garbage, delivered quicker than before.
A couple of weeks ago I wrote about building your own setup instead of buying another course. This is the other half of that letter. The setup is worthless if you aim it at the wrong link.
So here is the one question. Ask it before any tool, any automation, any purchase. Answer it without naming a single piece of technology:
Where does work wait the longest before it moves?
Not where people look busy, and not where it feels slow. Where the work itself sits, untouched, before someone picks it up and carries it forward. That is your constraint. That is where AI belongs. Everything else is expensive decoration.
It is the same instinct from that house in Conakry with no power, moved into a company with a budget. The instinct travels. The price tag is the only thing that changes.
This week: pick one process your team runs more than twice a week. Walk it end to end, slowly, and find where it stops. Then point your attention there, before you point your money anywhere. Diagnosis is still free.
Next week, the link nobody draws on the map: the people who have to use the thing you just bought.
Until next week,
Boubacar
I help solo consultants build AI operator stacks.
[email protected] · humanatwork.ai
Three things before you go
Forward this to one person who would use a kit on the Human at Work Shelf. One forward is more useful to me than ten retweets.
Was this forwarded to you? Subscribe at humanatwork.ai/atlas-sample. You also get the Mini-Atlas, three prompts I paste in my own work, free.
Reply with one sentence about what is stuck in your week. I read every one and it shapes what ships next.