Re-scoring AI readiness: theatre if nobody acted on the first score?

This piece closes the cluster this series has spent five pieces building, by circling back to where the whole programme actually started: the readiness score. Once an organisation has a first number, a specific temptation shows up. Scheduling a re-measure in ninety days feels like seriousness. It looks like tracking, like monitoring, like an organisation that takes its own governance seriously. But a second measurement is not automatically evidence of anything, and treating it as if it were is where re-scoring quietly turns into theatre.

The honest, contrarian version of this argument is simple: a re-score only means something if it is downstream of action. If nothing changed in the organisation between the first score and the second, the second number is not measuring progress. It is measuring whether someone remembered to run the survey again.

What makes a re-score legitimate

A re-score earns its value when specific findings from the first assessment led to specific interventions, a department got real attention, a red line got written into the policy, a training programme actually ran, and the second measurement tests whether those specific interventions worked. That is a real question with a real answer. "Time has passed, let's see if the number moved" is not the same question, and it does not have a real answer, because there is no mechanism connecting the two measurements beyond the calendar.

The theatre version, described plainly

The failure mode is familiar enough to name directly. An organisation runs the first assessment, gets its number, does relatively little in response beyond perhaps one training session, then re-runs the assessment roughly on schedule because ninety days appeared on a calendar somewhere. Whatever the second number shows, a small rise, a flat result, a small dip, gets read as meaningful, evidence that things are working, or evidence that more work is needed, without anyone being able to name the specific thing that changed and caused it. The number moves or does not move for reasons nobody can actually trace, which makes it noise dressed up as a metric.

The questions that separate a genuine re-score from theatre

• What specifically changed between the two measurements, tied to named findings from the first score, not a general sense that time has passed and effort has presumably been made.
• Which department or dimension was actually targeted, and whether the re-score is built to re-test that specific area rather than simply re-running the same generic survey end to end.
• Who owns the interventions that were supposed to happen in between, and can state plainly what they actually did.
• What the organisation will do differently if the second score comes back flat or worse, since a genuine measurement has to be capable of producing an uncomfortable answer, not just a reassuring one.

A second measurement is not evidence of progress. It is evidence that someone remembered to measure again. Only the gap between the two, explained by something specific that actually happened, is evidence of anything real.

When skipping the re-score is the more honest choice

If nothing was genuinely done in response to the first score, running the second assessment anyway is often less honest than simply saying so. A flat or similar number gets read as reassurance nothing has changed, badly. A number that moves gets read as progress the organisation cannot actually explain. Neither reading is true, and both are worse than the plain, uncomfortable admission that the interventions the first score called for have not happened yet, so there is nothing new to measure. The re-score should wait until there is a real intervention behind it worth testing, not arrive on schedule regardless.

The practical takeaway

Treat a re-score as legitimate only when it tests specific interventions that followed specific findings from the first measurement, and when nothing real has happened in between, say so plainly rather than running the assessment again and reading meaning into a number that has none behind it.

No external statistic cited; this article presents an internal contrarian argument rather than third-party evidence.