AI-Prepared Dashboards Add Data Faster Than People Can Decide

This paper argues that AI-prepared dashboards and digests add information faster than people can act on it, and proposes a Decisions & NBA Board: a short queue of decisions, each with cited evidence, options, an owner with authority and a dated check on the result. It matters because a report that changes no decision is a cost. No hypothesis has been tested.

A working paper: the claims are open to test and the text will change. Write to us for the current draft.

Abstract

AI can now connect to a company’s systems and assemble what people used to assemble by hand: the dashboard, the daily digest, the meeting brief and the Chief of Staff summary. That was a good place to start because it was new. It is no longer new, and it has a side effect. Every AI-prepared artifact adds to what a person must attend to, and the artifacts arrive faster than the person can act on them. Simon wrote in 1971 that a wealth of information creates a poverty of attention. If an AI-prepared dashboard or digest does not change what people decide, start, stop or measure, it is a reporting cost, and a better one is a bigger cost.

This paper proposes the layer above the dashboard: a Decisions & NBA Board. It is a short, continuously refreshed queue of decisions, each backed by cited operational truth, set against declared goals and guardrails, with options (including doing nothing), a recommended next best action, an owner with authority, and a dated check on the result. AI-prepared dashboards are its source of operational truth, and the Board does not replace them. The truth has to be built to be reproducible: derive each figure with a query where a query can produce it, add rules and a typed-decision model where queries run out, and use a language model only for what none of those can do.

Most of the paper is implementation. It gives a card format, rules for ranking without false precision, rules for what stays off the Board, a three-stage build (Crawl, Walk, Run) that needs no new platform, six worked examples, decision rights, failure conditions and a 30-day pilot. I also propose that a team stays deterministic as long as it can when it derives evidence, because a figure no one can derive again is not operational truth. I further propose that a team starts by sending the Board as a routine email to named readers, because the message forces the choice of the few items that matter and forces thought about who reads it. The principal hypothesis is that the Board improves target outcomes and shortens decision time more than an AI-prepared dashboard or digest of the same sources, without raising error rates. Eight hypotheses and a measurement design are given so others can show the claim wrong.