Scoring Method

ORAVORN uses a fixed scoring method. The same responses always produce the same result. Archetypes, domain readings and primary constraints are determined by the rules described on this page. A language model writes the narrative explanation from the computed result. It takes no part in the calculation.

Each scored question offers three levels, Low, Developing and Strong, and a fourth option, Not yet or Not applicable.

Not yet is used where a condition has not yet arisen, for example because AI has not been introduced. Not yet and Not applicable are left out of the calculation and are never treated as Low. Every result reports how many responses were scored and which areas remain untested.

A few questions offer named choices instead of levels, such as a strategic posture or a decision pattern. Each choice is assigned a reading of Low, Developing or Strong in advance.

The questionnaire feeds fifteen variables across the five domains of the SLGTC framework.

Some variables rest on two questions, and both responses count. Where both are needed, as with strategic clarity, the weaker response sets the reading. Where the two questions are different ways of seeing the same condition, matching responses give that reading. Responses one level apart give Developing, and so does a Low alongside a Strong, where the inconsistency is itself noted.

Some variables are worked out from combinations of responses. Competitive positioning compares the declared posture with what current AI activity is mainly for. A decision pattern marked by avoidance, deference or theatre cannot score Strong. Capacity relative to ambition cannot rate higher than strategic clarity.

Each domain's state is the median reading of its variables. When there is no single median, the more cautious of the two middle readings is used. The lowest variable in each domain is also recorded, so a significant weakness remains visible.

Each of the eight archetypes is defined by a set of characteristics across the five domains. Every characteristic is assessed as met, unmet or unknown. Some are more central to an archetype than others and carry more influence in the comparison.

The assigned archetype is the one whose characteristics are most fully met, judged by the share of its own characteristics that are unmet. A characteristic that cannot be verified is left out. One example is the outside pressure behind the Pressured Transformer, which depends on the respondent's own description.

Results report the assigned archetype, the quality of fit (Perfect, Good, Partial or Poor), the nearest alternative, and any close call. Where characteristics are unmet, the differences are described.

The primary constraint is the condition identified as the most useful starting point for action.

A small number of acute conditions take precedence because they combine elevated risk with active AI activity:

• AI deployed or scaling with consistently weak governance;
• AI scaling while both governance and change support are weak;
• AI activity under way without a specific strategic direction;
• AI activity under way with weak leadership alignment.

Where no acute condition applies, a sequencing rule is used. Where strategic clarity is Low or not yet considered, direction is identified first. A different domain takes priority only where it has failed structurally. Otherwise the weakest domain is identified as the primary constraint.

Where several conditions are failing, the calculation names one primary constraint and also reports the others that remain binding. No condition is assumed to make up for another. Fixed tie-breaking rules keep results consistent and repeatable.

The definition of a constraint and the principles behind the sequencing are set out in Constraints and Sequencing.

ORAVORN records differences between what respondents state and what their other responses imply. An organisation may describe its AI posture in one way while reporting activities that suggest another. Such differences are reported as findings in their own right. They do not change any score.

Free-text responses do not contribute to scores. They provide organisational context and specificity within the narrative explanation. Where they suggest a different interpretation from the structured responses, the difference is noted and the scores continue to reflect the structured responses.

The method does not adjust results for sector, organisation size or the reliability of individual claims.

The archetype, domain readings, fit assessment and primary constraint are derived directly from the responses provided and the rules described on this page.