Goal Engineering: scope agents around easy feedback
The point
Agents are moving toward open ended goals, LangChain says, so scope design matters. Pick a scope that makes it easy for a human expert to give feedback, because the context agents miss is often not written down.
Why it matters
For builders, scope decides how fast agents learn judgment only people carry.
Key details
- Actively AI switched from one agent per account to motions across many accounts
- Feedback on one motion task transfers to other tasks in that motion
- Background agents assemble account context for motion agents to use as needed
Original ~5 min → 25 sec here Source: LangChain Blog, Nov 14, 2026