HowtoConductanAIReadinessAssessment(The5-DimensionFramework)

An AI Readiness Assessment is the structured evaluation of an organisation's capacity to successfully design, build, and operate an AI product. In Business-First AI™, it is the primary output of the D2 Diagnose stage — and it is the most undervalued step in AI adoption.

**Why Most Organisations Skip It**

Readiness assessments feel like delay. Leadership wants to move fast. Engineers want to start building. The business analyst who proposes "let's assess readiness first" is often overruled by urgency and optimism. The consequence is that readiness problems are discovered at D5 or D6 — where fixes are 10x more expensive than they would have been at D2.

The AI Readiness Assessment does not gate AI adoption. It gates the current use case. An organisation scoring poorly on Data Quality for use case A can still pursue an AI initiative in a different area where data is stronger. The assessment tells you where to start — not whether to start.

**The 5-Dimension Framework**

Each dimension is scored 0–5. A score below 3 on any dimension is a blocker that should be addressed before D3 Define begins.

**Dimension 1: Data Quality (DQ)**

Score 5: Clean, labelled, well-documented data available for all target use cases. Data dictionary exists. Completeness rates above 95% for critical fields. Score 3: Data exists but requires significant cleaning and transformation. Known quality issues are documented. Score 1: Data is siloed, inconsistently formatted, or lacks coverage for the target problem. Score 0: Data does not exist or is inaccessible.

Key questions to answer: What is the completeness rate for each critical field? How frequently is the data updated? Is there a data dictionary? Have label quality issues been identified and quantified? Is the training data representative of the real-world distribution the model will encounter in production?

**Dimension 2: Process Maturity (PM)**

Score 5: All target processes are clearly documented, consistently executed, and measured. Process owners exist and are accountable. Score 3: Processes are understood but informally documented. Significant variation exists between individuals performing the same task. Score 1: Processes are inconsistent. The organisation cannot reliably describe what the current process is.

Key questions: Are the processes the AI will automate or augment documented in their current state? Is there a process owner? How much variation exists between individuals performing the same task? A model trained to automate an inconsistent process will be inconsistently useful.

**Dimension 3: Team Capability (TC)**

Score 5: Team includes ML engineers, data scientists, and BAs with AI product experience. Clear ownership of post-launch model maintenance exists. Score 3: Technical capability exists but AI-specific experience is limited. Dependency on external vendors is high. Score 1: No AI-specific capability in-house. No clear owner for model performance post-launch.

Key questions: Who will own the model post-launch? Who will investigate performance degradation? Is there a BA or PM with AI requirements experience? A capable ML team that ships a model and then moves to the next project is not an AI-ready team.

**Dimension 4: Leadership Alignment (LA)**

Score 5: Leadership has defined success metrics, explicitly accepted that AI outputs are probabilistic, committed to a measurement period before evaluating ROI, and approved the governance model. Score 3: Leadership is supportive but has not formally defined success criteria or accepted the implications of probabilistic outputs. Score 1: Leadership expects AI to behave like deterministic software — immediately perfect, requiring no ongoing maintenance.

This is the dimension most teams underestimate. Leadership who has not accepted that a model can be 90% accurate and still require refinement will interpret the first failure as a project failure. Setting this expectation is a prerequisite for a healthy AI product process.

**Dimension 5: Technology Infrastructure (TI)**

Score 5: Data infrastructure, compute resources, security controls, and compliance frameworks are in place for the target AI use case. Score 3: Infrastructure partially exists. Significant gaps require investment before the AI product can be deployed at production scale. Score 1: No AI-ready infrastructure. Starting from scratch.

Key questions: Can the data pipeline support the volume and velocity required for model training and inference? Are security and data privacy controls in place for the data the model will process? Can the existing systems accept the model's outputs?

**Interpreting the Results**

Sum the five scores for a total (max 25). As a general guide: 20–25 is green — proceed to D3 Define. 12–19 is amber — address the dimensions below 3 before proceeding. Below 12 is red — significant readiness investment required before any AI initiative.

The more actionable output is the Prioritised Blocker List: identify the 3–5 highest-priority gaps and assign an owner and a timeline to each. This becomes the first 90-day action plan before D3 begins.

Download the free AI Readiness Assessment template from the Business-First AI™ Framework Library to run this assessment in your organisation.