Field Note 002

AI theatre vs decision workflow change.

The work is not to add AI to every task. The work is to redesign the workflow so evidence, judgment, and accountability move together.

Most AI programmes are still judged by the wrong signals: tool usage, prompt volume, pilot count, assistant adoption, or whether a team can produce the same output faster. Those signals are not meaningless, but they are not transformation.

The harder question is whether the decision workflow has changed.

Has the path from evidence to recommendation to decision become clearer, faster, more trustworthy, and more reusable? Are sources visible? Are assumptions recorded? Does a human owner remain accountable? Do corrections improve the next run? If not, the organisation may have better tools but the same operating model.

The wrong unit: the task

AI can make individual tasks faster: summarising notes, drafting commentary, writing SQL, producing slides, or generating options. Useful, but insufficient. Faster tasks can still pile up behind unchanged approvals, unclear decision rights, weak evidence, and low trust.

The right unit: the workflow

The better unit is the analytics-led workflow: source gathering, context, analysis, interpretation, recommendation, review, decision, and follow-up. This is where AI either becomes practical infrastructure or remains theatre.

Five tests

  • Decision: Does this improve a real business decision?
  • Evidence: Are sources, assumptions, and caveats visible?
  • Allocation: Is it clear what humans, tools, and AI each own?
  • Trust: Is there review, accountability, and reversibility?
  • Learning: Does the workflow improve from corrections or outcomes?

What changes in practice

A good AI workflow does not simply draft a better readout. It creates an evidence layer, allocates work deliberately, captures a decision trace, adds a trust layer, and learns from human corrections.

Oakes & Signal view

Better decision workflows are the point. AI is an accelerant. The value comes when evidence, context, judgment, and accountability are designed together — especially in reporting, forecasting, measurement, planning, and executive readouts.

Open the Workflow Allocation Canvas See the framework