Dimension 02 · arqmetrica AI Maturity Index

Data foundations

The quality, governance, and accessibility of organisational data for AI use cases.

Weighted at 17% of the overall score

Framework anchors

The four questions in this dimension, with the four ordinal options and their fixed scores. Identical wording to what you will see in the assessment.

DAT-01

Think about the typical journey from "we want to try this" to "the model has the right data". Be honest, not flattering.

Option	Response	Score
a	Already in a clean, governed warehouse with documented schemas Hours to serve a request	100
b	Lives across 3–5 systems; needs joining each time Days to a couple of weeks per request	67
c	Quality is inconsistent; cleaned case-by-case Weeks to a month per request	33
d	We don't have reliable customer data for AI yet	0

DAT-02

Lineage = "for any field used by an AI model, can you trace it back to the source system, owner, and last refresh time?"

Option	Response	Score
a	Yes — automated lineage tooling, queryable in seconds	100
b	Documented for critical pipelines, manual elsewhere	67
c	Tribal knowledge — engineers know but it is not written	33
d	No lineage tracking	0

DAT-03

GDPR-relevant personal data has special obligations under the EU AI Act. This is about whether your systems know which fields are sensitive.

Option	Response	Score
a	Field-level tagging + automated DPIA workflow when AI touches PII	100
b	Manual DPIA process triggered by request	67
c	Reviewed only when legal escalates a concern	33
d	No formal handling of PII in AI workflows	0

DAT-04

Bias often comes from skewed source data. Have you formally measured representation gaps?

Option	Response	Score
a	Documented bias audit per major use case, refreshed quarterly	100
b	Spot-checks before deployment; no formal cadence	67
c	Aware of the risk but no formal process	33
d	Have not considered this	0

Take the full assessment to see your score across all six dimensions, your peer benchmark, and your three highest-leverage moves.