← sources

Everyone’s building agents. Is anyone actually good at telling them what to do?

source: reddit (reddit:AI_Agents)published 2026-11-07 07:17ingested 2026-11-08 02:35

retained text

I’ve spent the last few months watching people try to get AI agents to do real work, and the same thing keeps happening.

The agent is fine. The instruction is the problem. Something like “handle my outreach” or “research my competitors” goes in, and what comes back is generic, wrong, or confidently off in a direction nobody asked for.

The people who get good results seem to do something different before they start. They spell out the goal, what the agent can and can’t touch, and what “done” actually looks like. Most people skip that part, then blame the tool.

So I’m curious about your experience:

1. Have you handed a task to an agent and got back something unusable? What did you ask it?
2. When it worked, what did you do differently?
3. Is the hard part knowing what to ask for, or the setup (tools, permissions, connecting things)?
4. Would you pay for something that helped you write the brief, or is that a problem you’d rather solve yourself?

Not selling anything, just trying to work out whether this is a real gap or something I’m imagining.

document fetched 2026-11-08 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1xq0cb0/everyones_building_agents_is_anyone_actually_good