AI•DOOM Assessment: the first cut of an operating maturity diagnostic

The first AI•DOOM Assessment began with a management question: where would a change in working practice make a useful difference? Familiarity with AI tools was part of the answer. So were the way decisions moved, the quality of their inputs, and the amount of work that depended on one executive.

Michael E. Gruen developed the name during the initial exploration. AI•DOOM and its expansion, Artificial Intelligence Diagnostic of Operational Maturity, emerged together; the expansion was not retrofitted to a chosen word. The wordplay gave a serious subject a less solemn entrance. “Doom” was a name, not a forecast of what would happen to the person taking the assessment.

This essay describes that first version. The instrument has since changed, so its original dimensions and scores should be read as design history. The current assessment organizes its results around Leader Practice, Organization AI Maturity, and Operating Conditions. Those three facets remain separate rather than becoming one overall grade.

Why the first version started with work

A hypothetical executive can be comfortable using a model and still have a delegation problem. Drafts arrive quickly, but every exception waits for the executive’s approval. Another may use few tools personally while leading a team with clear ownership and dependable review. Tool familiarity alone would tell an incomplete story about either situation.

The original assessment tried to make that distinction visible through sixteen questions across eight dimensions. Four concerned operating practice: judgment, decision velocity, communication leverage, and delegation capacity. Four concerned AI practice: harness proficiency, prompt engineering, reverse-prompting proficiency, and organizational impact awareness.

Some of those labels needed explanation. “Harness proficiency,” for example, concerned the arrangement around a model: context, tools, handoffs, review, and stopping conditions. The question was whether the work had a usable method, beyond a prompt that happened to produce a good answer once.

Two questions per dimension could provide a starting point for discussion. They could not establish how reliably an executive or organization would perform under different conditions. The original summary scores, including the Operating Leverage Score and AI readiness to leverage, compressed those answers further. Their apparent precision should never have been mistaken for stronger evidence.

What a result should lead to

A useful readout gives the executive a question worth investigating. If communication depends on one person drafting everything, inspect one recurring message. If decisions keep returning for approval, look at the decision rights and the information available to the team. If an AI-assisted task works only when its creator is present, ask what another person would need to repeat it.

These are examples of follow-up work, not conclusions that a questionnaire can establish on its own. The executive’s answers need to meet actual records, observed practice, and the perspectives of people who do the work. A discrepancy is useful material for the conversation.

The first version’s purpose was to make a broad subject concrete enough to examine. Its limits also gave the next version something to improve: clearer distinctions, more attention to evidence, and less temptation to read a single number as an identity.

The name can open the conversation with a little humor. The instrument earns its place only if the person leaves with a more precise question than the one they arrived with.