I built free AI agents for WordPress — here's what the first video taught me
Six months in building Agent Builder, a free WordPress plugin that lets you deploy AI agents using plain job descriptions instead of prompts.
Just published the first walkthrough video and the thing that surprised me: the most common question in comments wasn't "how does it work?" — it was "can it do X without me reviewing every output?" Every non-technical user's first instinct is full autonomy. Every technical user's first instinct is the opposite.
We ended up shipping a built-in approval workflow as the default, with autonomy as opt-in. That one decision changed how we explain the product entirely.
If you're building anything with AI automation — especially for non-technical users — curious whether you've hit the same autonomy vs. control tension, and how you resolved it.
Video if useful: https://www.youtube.com/watch?v=sMCnXQGb3nk
Product page: https://www.indiehackers.com/product/agent-builder-2
The split between non-technical users wanting autonomy and technical users wanting control is interesting. It sounds like the approval workflow didn't just solve a product concern—it gave you a much clearer default for how the product should be trusted. The fact that this emerged from watching what people asked after seeing the walkthrough is probably the bigger signal.
The autonomy-vs-control discovery is the real product insight here. Most founders building AI tooling assume users want maximum automation, but your first video revealed that non-technical users are actually asking "will this do things without my permission?" That's not a friction point - it's a feature requirement. The fact that approval workflows became your default design means you're building for trust first, then adding autonomy as an opt-in for users who've already verified the agent's judgment. That's the right order. Most automation tools get it backwards.