As AI agents gain authority to act on behalf of companies, governance must define who owns their decisions, where autonomy ends and how permissions change over time, writes Humanity founder Terence Kwok.