OKR planning as belief revision

Objective and Key Results (OKR) is a well-known framework used to set and track an organization's goals. Leaders establish top-level OKRs, which then flow through the organization. In this journey, more details are added, these strategic goals drive alignment and but also adjust according to tactical needs and constraints. Progress is monitored, and any necessary changes are made throughout the organization.

Large organizations are structured in complex hierarchies because this setup helps with planning and execution, compared to other options.

The concept of belief revision has been explored in various fields for a long time. In 1982, Judea Pearl introduced belief propagation, an algorithm designed for making predictions using graphical models like Bayesian networks or Markov random fields. When applied to a tree-shaped model, this algorithm finishes in just two complete cycles. On more complex graphs, an approximate version of belief propagation can be used, though it requires more steps – which is costly when implemented at the speed dictated by human processes.

AI and LLM will make these processes faster and lower coordination costs.

There are many other functions and factors that stabilize bureaucracies, but they are ultimately bound by what they achieve. The often unstated purpose of large corporations is not to compete in the market and make profits, it is to grow beyond that game and "engulf enough of the world to shield themselves from uncertainty" (rephrasing cit. John Kenneth Galbraith and Danella Meadows's Leverage Points). When they are successful for a long time, they can loose the ingredients and factors that made them good at the innovation game in their earlier history. Increasing profits is a necessary condition to continue to play. Inefficiency and slow speed are a moral hazard and mortal sin for a corporation that faces market shakeups and looses mindshare even before marketshare.