Methodology

How we map the agentic data journey

One readiness score, six pillars, and a transparent set of visible signals. The benchmark estimates whether an organization has the conditions to move from distributed AI experimentation to governed scale.

One score: readiness for governed agentic scale

Every organization receives a score from 0 to 100. The score does not measure AI enthusiasm or model access alone. It measures the visible operating pattern around agentic data work: usable data, a build layer, retrieval, evaluation, governance, hiring momentum, and leadership ownership.

Treat the result as a directional benchmark. It is reconstructed from observable signals and will not capture every proprietary system or internal workflow.

Five stages of the journey

Score
Stage
Description
0–19
Not ready
Visible signals are still limited. The organization may have isolated data tooling, but the foundation for reusable agentic workflows is not yet apparent.
20–39
Foundation
A data foundation is forming. The next step is to connect modern data infrastructure with retrieval, agentic construction, and clearer ownership.
40–59
Experimenting
AI workflow signals are visible, but the operating pattern is incomplete. Context, retrieval, evaluation, governance, or leadership may still be fragmented.
60–79
Operationalizing
The organization shows meaningful capability and momentum. The challenge shifts from isolated builds to repeatable, governed workflows across teams.
80–100
AI mature / scaled
The visible stack is production-shaped: modern data, agentic build capacity, evaluation, governance, momentum, and ownership reinforce one another.

Six pillars of governed agentic scale

How to interpret the score

The benchmark is strongest as a directional map of enterprise adoption. Frontier AI producers like OpenAI or Anthropic may build equivalent systems internally, making their maturity less visible through third-party technology signals alone. In addition, some labs obscure their titles via umbrella roles such as 'Member of Technical Staff'. Read each score alongside the organization's archetype the evidence we have used to score it.

Score constraints

The model applies constraints when a high score would overstate visible maturity. Examples include a missing modern data stack, a heavily MDS-skewed stack with limited AI-native tooling, or an absence of evaluation, observability, and leadership signals. A total absence of LLM, tooling, job-description, or project evidence also creates a negative adjustment.

These constraints help distinguish analytics maturity from production-shaped AI readiness.