How the organisation's data supports predictive analytics, AI, and machine learning, including transparency, bias mitigation, and model governance.
INT-01
Enterprise AI strategy and portfolio governance
Does the organisation have an executive-endorsed AI strategy governing which predictive analytics and Decision Intelligence initiatives are pursued, with a managed portfolio view rather than uncoordinated, team-level experimentation?
Maturity level descriptions
No AI strategy exists; predictive analytics and Decision Intelligence initiatives, where they occur, happen in an uncoordinated way with no executive oversight. Individual teams experiment with predictive or AI-driven tools independently, with no enterprise strategy or executive visibility.
Awareness exists that AI strategy is needed, but nothing has been formally endorsed, and initiatives remain uncoordinated across the organisation. Senior stakeholders acknowledge AI strategy is needed, without a formally endorsed strategy or a consolidated view of current initiatives.
An AI strategy is documented and endorsed by the executive team, with a defined process for approving new AI/predictive analytics initiatives. A documented, endorsed AI strategy sets out priorities and a defined approval process for new initiatives.
AI initiatives are tracked as a managed portfolio against the strategy, with value, risk, and resourcing reviewed by the executive team on a defined cycle. A managed AI portfolio is reviewed on a defined cycle, with value, risk, and resourcing decisions made at executive level.
AI strategy and portfolio decisions are dynamically adapted to emerging capability and risk, with board-level oversight and external benchmarking of the organisation's AI maturity and value realisation. The AI portfolio is actively reprioritised as capability and risk evolve, with board-level oversight and external benchmarking of AI maturity and realised value.
INT-02
AI model lifecycle and governance framework
Is there an executive-endorsed model governance framework covering the full AI model lifecycle (development, validation, deployment, monitoring, retirement), and does the organisation know how many models are actually in production against it?
Maturity level descriptions
No model governance framework exists, and the organisation does not know how many AI/ML models are in production or what they do. Models can be built and deployed with no reference to any governance framework, and no central record of what is in production exists.
Awareness exists that model governance is needed, and some models are known informally, but there is no formal framework or comprehensive inventory. Some models are known about informally by those who built them, without a formal governance framework or comprehensive inventory.
A model governance framework is documented and endorsed, with a defined risk classification and approval process, applied to priority AI initiatives. A documented, endorsed framework defines risk classification and approval requirements, applied to at least priority AI initiatives.
All models in production are covered by the governance framework and tracked in a comprehensive inventory, with lifecycle status reported to the executive team on a defined cycle. A comprehensive model inventory covers all production models, tracked against governance framework requirements, with status reported to the executive team.
Model governance is enforced technically within the model development and deployment pipeline, with real-time, board-level visibility of the organisation's full AI model portfolio and its governance status. Governance requirements are technically embedded in deployment pipelines (e.g. models cannot go live without passing governance checks), with real-time board-level visibility of the full model portfolio.
INT-03
Algorithm transparency and bias mitigation as an enterprise ethics and risk accountability
Does the executive team treat algorithm transparency and bias mitigation as an owned, resourced enterprise risk and ethics responsibility, with clear accountability, rather than a technical concern left entirely to individual data science teams?
Maturity level descriptions
No executive-level accountability exists for algorithm transparency or bias mitigation; these are treated, if at all, as purely technical concerns for individual teams. Bias and transparency concerns, where they arise, are handled entirely at the technical team level with no executive ownership or resourcing.
Awareness exists that bias and transparency are organisational risk issues, but no executive owner or dedicated resourcing has been assigned. Senior stakeholders may express concern about bias or transparency risk without assigning clear ownership or resourcing to address it.
A defined executive-level accountability (e.g. a nominated risk or ethics owner) exists for algorithm transparency and bias mitigation, with a documented approach applied to priority AI initiatives. A nominated executive or senior role owns transparency/bias risk, with a documented approach applied to at least priority AI initiatives.
Bias mitigation and transparency requirements are systematically applied and tracked across all material AI initiatives, with findings reported to the executive team or board. Comprehensive application of bias mitigation and transparency requirements is tracked across all material AI initiatives, with findings formally reported at executive level.
Bias and transparency risk is integrated into the organisation's broader enterprise risk and ethics framework, with independent assurance (internal or external) and public or regulatory reporting where relevant. Bias and transparency risk is managed alongside other enterprise risks, with independent assurance processes and, where relevant, public or regulatory disclosure.
INT-04
Enterprise AI value realisation and return on investment
How well can the organisation demonstrate, with evidence, the actual business value realised from its AI and advanced analytics investments, as opposed to relying on the promise of future value?
Maturity level descriptions
No measurement exists of value realised from AI investments; initiatives proceed on the assumption of future benefit with no evidence collected. AI initiatives are funded and continued based on anticipated benefit, with no mechanism to measure whether realised value actually materialises.
Value is occasionally estimated informally for specific high-profile initiatives, but there is no consistent measurement approach. A specific initiative may have an informal value estimate produced for a particular occasion, without a repeatable measurement method.
A defined value measurement approach exists and is applied to priority AI initiatives, with actual value tracked against initial business case projections. Priority AI initiatives are measured against their original business case using a defined, consistent value measurement approach.
Realised value is tracked across the full AI portfolio, with underperforming initiatives identified and reviewed, and results reported to the executive team. Portfolio-wide value tracking identifies underperforming initiatives for review, with consolidated results reported to the executive team.
AI value realisation is a standing input to investment decisions, with realised value reinvested into further AI capability and benchmarked against industry value-realisation standards. Demonstrated AI value directly shapes ongoing investment decisions and reinvestment, with the organisation's value realisation benchmarked against industry standards.