01
Handed to me
Teams that already paid for ChatGPT and Claude, and still worked exactly the old way.
MUCH · Buenos Aires · Aug – Dec 2025
They had ChatGPT. They still worked the old way.
As an AI consultant, I automated clients’ day‑to‑day work: meeting note-takers, custom CRMs, small tools for one job each. I also ran adoption end to end: shadowed people first, built around how they actually worked, and trained them until the new way stuck.
AI consultant30%+ less work time≈ 50% more output
Shadowing · one workday6 h of tasks
What changeslive
0.0 h back a day, a month later
On paper: 0.0 h0% of her day
Some tools save more on paper than they ever will in practice.
Sample data, made up for this page: the person, her day and every minute. It is not a client’s workday, and the minutes saved are not the results from the real work.
What the five months looked like
Four parts, in the order they happened.
02
Noticed
The tools weren’t the problem. I shadowed employees first and mapped how their work actually flowed.
03
Built
Meeting note-takers, custom CRMs and task-specific tools built around those patterns, or existing tools adapted to them.
04
Changed
−30%+
work time, and output up about 50%.
The rest of the job
What sat around the tools.
- Shadowing firstSat with employees to map how work really flowed before building anything.
- Asking whyDiagnosed why teams with ChatGPT and Claude subscriptions stayed in their old workflows, and ran adoption surveys.
- Simpler toolsBuilt AI systems around the patterns people already had, or adapted the tools they already used, and shipped simpler ones where needed.
- Making it stickSOPs and hands-on training, end to end, until the new way stuck.
Honest limit: client work, so no client names or screens here. The percentages come from those engagements, not from this demo; the workday above is invented.