Accounting in the Age of AI: Rebuilding Trust, Reducing Friction, and Raising Economic Quality

Repository abstract

In the present AI transition, the accountability dividend is the hypothesis that more reliable, usable and contestable accounting can release productive capacity by reducing avoidable administrative work. This paper treats that dividend as a proposition to test, not as an established national effect. Administrative reform can improve entrepreneurial opportunity, but evaluating it requires separating production from turnover, professional responsibility from outsourced work, and simulation from causal evidence. This study investigates Portugal through an archived Eurostat economic baseline, a measurement audit of startup claims, selected national legal sources, and six reproducible computational experiments. A comparison of Portuguese startup turnover with United States small-business GDP cannot establish a startup contribution gap. An exact broad-sector decomposition of nominal value added per hour finds that within-sector differences dominate the Portugal–EU level difference in the accessed data; this is an accounting result, not a purchasing-power-adjusted explanation of productivity. Existing invoicing and reconciliation systems weaken the premise that integrated transaction infrastructure would be wholly novel. Portuguese provisions distinguish organised accounting and reserved technical responsibility from a universal requirement to outsource routine tasks. Synthetic models expose tradeoffs: removing fees can increase founder work or rework, and reducing queues does not guarantee better cash survival once implementation costs are included. No Portugal-wide ten-percent-of-GDP administrative loss or causal accounting-regulation effect is identified. The defensible policy contribution is an interoperable evidence architecture, a controlled comparison of self-management and targeted review, and an evaluation framework that retains integrity, privacy and appeal safeguards. Representative prices, satisfaction, authenticated portal performance, detailed EU-27 reserved-act rules and national causal effects remain incompletely established.