I develop a principal-agent framework in which agents take actions that influence a firm's general ledger accounts. My framework is tractable and facilitates optimal closed-form solutions without assuming preferred actions or imposing exogenous restrictions on contractual form. I apply the framework to several settings and a consistent theme emerges: it is cheaper to motivate efficiency improvements than revenue growth. This is driven by a seemingly trivial property of bookkeeping, that general ledger accounts are bounded below by zero. I find that this accounting feature influences task allocation, organizational design and optimal aggregation rules. My analyses produce many readily-testable predictions and can help explain empirical regularities; for example, I predict that the low pay-performance sensitivities empirically observed in loss firms may be driven in part by life cycle. My findings demonstrate that when studying the optimal use of accounting information, properties of the double-entry systems which generate that information should be taken seriously.