4 memos. klarna 2024, duolingo 2025, IBM 2025, intuit this tuesday. 3,000 jobs in the latest one. every memo names what got cut, what got refocused, what stays. zero of them name the human who answers when the AI is wrong.
i think solo builders feel this gap first because we replaced contractor work or hires with agents already - the cs email, the lead-qual script, the invoice categorizer, the refund decision flow. when the agent ships something wrong (wrong tone, wrong refund, wrong category, wrong promise to a customer), the failure path is just... us, by default, because we didn't write the line.
intuit's memo is the same gap at 3,000x scale. that's the part that scares me. enterprise PMs are gonna inherit the unstaffed accountability when something breaks under stress instead of authoring it on purpose now.
honestly curious how other solo builders handle this. when your agent gets the work wrong, is there a written line - a doc, a note, a one-pager - that says "this one is on me"? or is it just implicit, defaulted, the thing you only think about after a customer complaint lands?