A next action is the single visible move that gets motion back into the work, not the whole project and not the outcome you want.
Think about clearing a jammed printer. “Fix printer” is a label for the whole problem, not a next action. “Open the paper tray and remove the stuck page” is a next action: you can see it, you can do it, and after that the next action may change. “Launch the client portal” is a goal. “Email Sam for the DNS login” is the next action.
How it shows up
In GTD, captures land in an inbox, then you process them, and the main question is what the next action is. If there’s none, maybe it’s reference, trash, or someday maybe. If it belongs to someone else, it becomes waiting on. If it belongs to you, it needs a verb and enough context to do later without rethinking everything. This is why vague task lists feel heavy. “Client onboarding” just sits there and makes you feel behind. “Send the onboarding folder link to the client” is lighter because it names the move. AI can turn messy notes into candidate next actions, but you still need judgment: it may suggest ten actions when one matters, or miss who owns the step.
Why you care
Most stuck work isn’t stuck for lack of ambition. It’s stuck because the next visible move is unclear, so people reread the same email and reopen the same note without doing anything. A good next action removes that reload cost and makes delegation cleaner. “Help with the portal” names no action. “Confirm which domain the portal should live on” creates movement and ownership. For agentic work, concrete next actions keep AI grounded: it makes progress on a clear step and struggles with a foggy wish.