Our free assessment tools, the AI Readiness Assessment and the WC Self-Assessment, invite readers to answer three optional questions about their sector, organisation size, and role level before seeing their results. Those answers, together with anonymous scores, accumulate into the dataset behind any benchmark we later publish, including an eventual annual State of AI Readiness report. This page sets out the publication rules in advance so that when a figure appears, you can see exactly what standard it had to meet.
Sample thresholds
A benchmark cut is only reported once it reaches a minimum sample. The thresholds are:
- 30 completions per demographic cut. Any figure sliced by sector, organisation size, or role level (or a combination) must rest on at least 30 completed assessments in that cut.
- 100 completions per role for role benchmarks. The existing rule for the role-level averages shown inside the tools is retained: those stay labelled illustrative until a role has 100 or more real completions.
A cut below its threshold is never reported. It is not reported with a caveat, an asterisk, or a wide error band; it simply does not appear. Small cells are also a privacy protection: a threshold of 30 means no published figure can describe a group small enough for anyone to guess who is in it.
Who the data comes from
The data comes from a self-selected panel: readers of this site, mostly Australian practitioners, who chose to complete a free assessment and chose to answer the optional questions. It is not a representative population survey and we will never present it as one. People who find an AI-education site and finish an AI readiness assessment are, by definition, more engaged with AI than the workforce at large, so the panel likely skews toward higher awareness and interest.
That self-selection limitation will be stated prominently wherever benchmark figures are published. The honest claim these benchmarks can support is "among practitioners who assessed themselves on TheAICommand", and that is the claim we will make.
Aggregates only
Only aggregates are ever published: averages, distributions, and counts across groups that meet the sample thresholds above. Row-level data, meaning any individual assessment result, is never published, shared, or sold, in any form. There is no personal identifier in the data that could link a row to a person (see the data model below), and the aggregate-only rule ensures no published figure narrows down to an individual either.
What is stored
In plain terms, a completed assessment stores one anonymous record containing:
- which tool was completed, and which role path or version was taken;
- the numeric dimension scores and the overall result tier;
- the three optional demographic answers, stored as fixed category codes (for example,
governmentorsize-20-199), each with an explicit prefer-not-to-say option. A skipped question stores nothing at all; - a timestamp.
There are no personal identifiers of any kind: no name, no email address, no IP address, no device fingerprint, and no free text. The questions are fixed multiple choice, so nothing you could type ever reaches the dataset. Storage is insert-only: records are written once and cannot be read back, edited, or enumerated from the website. Answering the demographic questions is always optional, and skipping them never delays or blocks your results. Collection only happens with analytics consent; see the Privacy Policy for the full disclosure.
Current status
No benchmark report has been published yet. Collection of the optional demographic categories began in August 2026, and the benchmarks currently shown inside the tools remain synthetic and labelled illustrative until the thresholds above are met. When the first report is published, it will link back to this page, and any change to these rules will be dated here before it applies to a published figure.
General information and education only. Not legal, compliance, financial, or professional advice.