AI agents really try before saying they could not
- An AI agent could tell someone it could not add something to a sheet, or send a message, without ever trying, and then repeat that in later replies to the same person. Now an agent that says it did, or could not do, what one of its tools does, without having used that tool, is sent back once to really use it, and the run says so.
- A tool that adds, saves or sends is now used whenever your instructions say to (for example, add every opportunity to the tracker), not only when the person asks for it.
BlocksAI agent