Everyone seems focused on building more AI agents rn. But I've been thinking about what happens a year or two later.
Different teams build agents for different workflows. Some end up doing almost the same thing. Some stop getting used. Some still exist even though the process they were built for has changed.
We've seen this happen with internal tools, scripts, and even microservices. They solved real problems at the time, but very few teams were excited about cleaning them up later.
I wouldn't be surprised if AI agents end up following the same pattern.
Has anyone started thinking about this already, or do you think better governance and agent platforms will keep it from becoming a problem?
[link] [comments]