Right now Claude Code is your thinking layer and Clay is your execution layer because you don't trust Claude Code to write to client systems yet. But that boundary is going to move — probably faster than people expect. The interesting question is what your stack looks like the day you do trust it to write. Does Clay stay because it's the right execution layer, or because it was the execution layer when you built the workflow?
Nice — CRM cleanup is a good first crossing. Lower stakes than writing into a sequence, but it's the same trust gradient. Once you've verified it doesn't break anything, the next "I won't let it do that yet" loosens up pretty quickly.
The read/write split is the part I'd push on.
Right now Claude Code is your thinking layer and Clay is your execution layer because you don't trust Claude Code to write to client systems yet. But that boundary is going to move — probably faster than people expect. The interesting question is what your stack looks like the day you do trust it to write. Does Clay stay because it's the right execution layer, or because it was the execution layer when you built the workflow?
Good point, this is already outdated, I cleaned the CRM of one of my customers just using Claude code.
Nice — CRM cleanup is a good first crossing. Lower stakes than writing into a sequence, but it's the same trust gradient. Once you've verified it doesn't break anything, the next "I won't let it do that yet" loosens up pretty quickly.
insightful and gave me great ideas!
top ! je ferais des vidéos aussi car sur Claude Code ça avance vite ! tu vas bien ?