Job Hatred Is the Agent Opportunity Filter
Maps to joelclaw agent-loop intake: find hated work where current AI can execute and the business already feels the pain.
Malte Ubl framed agent discovery with one beautifully blunt question: “What do you hate most about your job?” Not “where can we use AI?” Not “what would make a slick demo?” What do people already resent doing?
That question is useful because it filters for pain before architecture. The attached Vercel stage slide says the low-hanging fruit for agents sits where current-gen AI can actually solve it and the work has high business impact. That’s the sweet spot: not speculative sci-fi, not toy automation, but hated work with receipts.
For joelclaw, this is a cleaner intake question for agent loops and internal workflow design. Start from the human complaint, then test whether the task is structured enough for an agent, valuable enough to matter, and annoying enough that people will actually adopt the fix.
Key Ideas
- Malte Ubl uses “what do you hate most about your job?” as a discovery question for deciding which AI agents to build.
- The opportunity filter is two-part: the job has to be solvable by current generative AI, and it has to carry real business value.
- Hated work is stronger signal than abstract “AI use cases” because it points at existing friction, budget, and urgency.
- For joelclaw, the pattern fits agent-loop scoping: capture the complaint, define the boring repeatable work, then decide if it deserves automation.