Agentic AI value lies in infra, data and change management, not building the agent

alex_verem · x · 2026-11-14

Mardehaym argues that anyone can build an agent over a weekend, but enterprise value goes far beyond that:

Infrastructure first: connecting systems, setting permissions, and making sure someone can audit what the agent did and step in when things go wrong.
Data is the core: where does it live, is it current, and do two departments mean the same thing by "revenue"? Agents need these questions answered to do useful work.
Change management must run alongside the build: people need to know what changes in their day-to-day work, what they remain responsible for, and when to take over.

alexverem calls this one of the most underrated agentic AI accounts on X.

More from coding & agent