Two experienced architects get the same job: find and map an ancient city hidden in the desert.
Jeff’s way
Jeff drives straight into the desert.
He meets a man in the dunes and asks for directions. The man points. “That way, about forty minutes.”
Jeff drives. Forty minutes pass. Nothing but sand.
He spots someone else far away, drives over and asks again. This man says Jeff has gone past the entrance and points back the other way.
Jeff drives until dark, sleeps in the car, and in the morning finds tyre tracks. He follows them to the city.
Expected: one hour. Actual: one day.
At the entrance he hires a guide for $100 and maps on foot with his tablet and camera. They cover one corner of the city. The guide says that’s it. Jeff thinks it was a good day’s work. He has mapped about a sixth of the city.
Hotel, sleep, repeat.
Seven days. $1,300.
Steve’s way
Steve arrives at the edge of the desert with a small drone.
Before driving anywhere, he flies the drone and maps the area from above. He sees the city straight away and drives to the entrance in 90 minutes.
He does not hire a guide. He uses the aerial map to pick the important areas, then explores them on a small electric bike, one area at a time.
A day and eight hours. $700.
It was not the drone
It is easy to say Steve won because he had a better gadget.
But Jeff was not short on gear. He had a car, good binoculars, a tablet and cameras.
The real difference was three habits:
Steve got the full picture before committing. He looked at the whole area first. Jeff drove in and started guessing.
Steve did things in a sensible order. Wide view first, close detail second. Jeff kept repeating the same slow work without knowing how much was left.
Steve did not trust directions blindly. Jeff took a stranger’s word twice and lost hours both times.
What this means for AI
Buying more AI tools, subscriptions and plugins without a plan gives you Jeff’s result at a higher price.
To work like Steve:
- Look at everything first. Before picking what to research in depth, look at the whole pile: tickets, logs, reviews, overall numbers.
- Go wide, then deep. Use a quick, broad AI pass to find the main patterns. Then run focused interviews on those patterns.
- Check every pointer. An AI suggestion is a stranger in the dunes pointing somewhere. It might be right. Check before you drive.
The one line to remember
Jeff had tools. Steve had a process. Build the process before you buy the gear.
Sometimes the right process means running AI on your own machine. That is next: why you might want your own AI setup.