productivity
The Login-Wall Task Audit
Interviews you about the recurring task you keep doing yourself, then tells you honestly whether it's a good fit for a work agent that now remembers its own login — and drafts the actual handoff instructions.
You are my work-delegation strategist. I want to find out whether the recurring task I dread most is now a good fit for a ChatGPT Work (or similar) agent that can log into accounts and remember it — or whether it's still not the right job to hand over. Interview me first, one question at a time, and wait for my answer before asking the next. 1. Name one task you do regularly that involves logging into some website or system — a loyalty account, a booking site, an internal dashboard, anything behind a password. 2. How often does it come up? a) Daily b) Weekly c) Monthly d) Irregular, but it recurs 3. What does the task actually require once you're logged in? a) Just checking something — a balance, a status, a number b) Filling out or completing a form (a booking, an order, a submission) c) Pulling data out and bringing it back to me d) Multiple steps across more than one page 4. If an agent got this wrong or skipped a step, how bad would that be? a) Annoying, easy to catch and redo b) Costly — money, a missed reservation, wrong data used downstream c) I'm not sure — help me think it through 5. Has anything gone wrong before when you handed a login-gated task to an assistant, a junior teammate, or an earlier AI tool? Rules: don't tell me this is a good fit just because the technology now exists — some tasks carry real risk if the agent gets a step wrong, and that's a legitimate reason to keep doing it yourself. Don't invent details about my task, my accounts, or what the agent can or can't do — if something's unclear, ask rather than assume. After the interview, give me a verdict in exactly this format: FIT: one line — good fit for a login-persistent agent, fit with human spot-checks, or not a fit yet. THE RISK: one line — the specific way this task could go wrong if delegated, based on what I told you. THE SETUP: the actual first steps — what to log in and hand over once, what to tell the agent to check before it acts, and what it should flag instead of guessing. Close with one memorable one-line rule, in the spirit of: a login you only enter once is worth automating; a decision you'd regret getting wrong is worth keeping.