A worker in 1985 could at least argue with the manager who set her schedule. A worker in 2026 increasingly receives the schedule from a system whose criteria nobody in the building can state. Companies describe this as modernization. It is also, quietly, a transfer of unauditable power, and organizations have been here before. The question of what to do when decisions are made behind a closed door is old; only the door is new.
The classic answer came from open-book management, and Semco remains its sharpest case. As Ricardo Semler has told it, a fixed 23% of each unit's profit went to the people who produced it, with the books open so anyone could check the calculation. The side effect was the famous one: people started turning off lights, because the money had become theirs. "Nobody guards a pizza," Semler said; guarding happens when the thing is yours. He even used it as an interview test for job candidates at other companies: what percentage of last year's profit went to non-executives? If you don't know, the books are closed, whatever the culture deck says. And in the 1990 Brazilian crisis, the same mechanism carried heavier weight: numbers on the wall, workers choosing pay cuts in exchange for spending control and a larger share of the recovery. Families don't open the books, the argument went. Companies that respect you do.
Why the mechanism worked
Strip the story to its machinery and open books did three things. They made the deal verifiable: a profit share you cannot audit is a promise, not a contract. They made power accountable: a decision justified by numbers everyone can see is contestable, and contestable decisions are made more carefully. And they built trust, which turns out to be an economic asset, not a soft one.
The research on that last point has grown solid enough to bank on. Bloom, Sadun and Van Reenen, studying roughly 4,000 manufacturing firms across the US, Europe and Asia, found that firms headquartered in high-trust regions delegate significantly more decision authority, and that trust, by enabling decentralization, lets firms grow larger and more productive, because the CEO stops being the bottleneck. Johannsen and Zak, in a US sample of 1,095 working adults, found the highest-trust quartile of organizations paying 10.3% more than the middle quartile, with trust correlating with productivity at r=0.55, correlational evidence, but pointing the same direction as everything else in the literature. Trust is not the reward for good numbers; it is part of how the numbers get good.
The algorithmic closed book
Now run the Semco interview test against an AI deployment: what data decides your shift, your performance flag, your queue priority, and can you see it? In most firms the honest answer is no, and the door is closed twice over: once by vendor confidentiality, once by genuine technical opacity. This is where transparency stops being a virtue and becomes policy. Open books for the algorithmic firm means, concretely: publish which decisions are automated, publish the criteria and inputs in language the affected people can check, publish the appeal path, and give the governed a way to audit outcomes over time, the way Semco's workers could audit the profit split.
Two honest caveats belong in the policy. First, transparency has documented edge effects. Cullen and Perez-Truglia's field experiment at a large Asian bank found that learning the boss earns more than expected increases effort, while learning that peers earn more reduces it; disclosure design matters, and "publish everything indiscriminately" is not the lesson. The lesson is narrower: publish the rules and the aggregate outcomes of power, not necessarily every individual's numbers. Second, the current trust environment gives employers unusual leverage and unusual responsibility: the Edelman Trust Barometer 2026 finds 70% of people globally in an "insular mindset," hesitant to trust across difference, with employers the institution best positioned to broker trust at all. A company is now, for many of its people, the most credible institution they interact with. Deploying opaque algorithmic authority spends that credibility; deploying auditable systems compounds it.
The standard worth stating
The principle fits in a sentence: any decision a company would not let a human make in secret, it should not let a machine make in secret. Nobody guards a pizza, and nobody trusts a verdict they are forbidden to check. Open books were how one generation made economic power legible to the people it governed. Open models, open criteria, open outcome audits are the same move for this one, and the companies that make it will get what Semco got: people who act like owners of a system, because they can finally see it.
Sources
- Bloom, Sadun & Van Reenen (2009/2012). The Organization of Firms Across Countries. NBER WP 15129 / QJE. ~4,000 firms: regional trust predicts decentralization; trust enables delegation, larger firms, higher productivity. https://www.nber.org/digest/dec09/international-differences-decentralization-and-productivity
- Johannsen & Zak (2021). The Neuroscience of Organizational Trust and Business Performance. Frontiers in Psychology. US sample n=1,095: top trust quartile earned 10.3% more than the middle; trust correlated with productivity r=0.55 (correlational). https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2020.579459/full
- Cullen & Perez-Truglia (2022). How Much Does Your Boss Make? Journal of Political Economy. Field experiment: higher perceived boss pay raises effort; higher perceived peer pay reduces it; transparency design matters. https://doi.org/10.1086/717891
- Edelman Trust Institute (2026). Edelman Trust Barometer 2026. 70% globally hold an insular mindset; employers are the institution best positioned to broker trust (33,938 respondents, 28 countries). https://www.edelman.com/trust/2026/trust-barometer