The AI Company Data Initiative (AICDI) is introducing a new scoring methodology for 2026, and we would like you to join the conversation and contribute your insights to this important development. This represents a shift from assessing whether companies disclose information on AI governance to assessing the quality, maturity, and evidentiary strength of the practices behind that disclosure, and this consultation is your opportunity to help shape that framework before it is finalised.
Why are we refining the methodology?
The AI Company Data Initiative (AICDI) is currently in its three-year pilot phase, during which we are actively testing, learning, and refining our approach based on real-world experience. Over the course of this initial phase, our ongoing engagement with companies, investors, and other stakeholders has surfaced a consistent theme i.e. an interest not only in whether companies disclose information on their AI governance practices, but also in understanding the maturity, quality, and evidentiary strength of the practices that sit behind those disclosures.
Building on transparency as an essential foundation, the refined scoring methodology introduces a more granular lens, allowing it to distinguish between companies that state intent and those that can demonstrate evidenced, working governance practice. This document sets out how company responses to the AICDI survey are assessed and translated into comparable scores across all 11 AICDI modules.
This consultation period starts on 24 August and is open till 4 September 2026.
The Guiding Principles
The redesign is anchored in one core idea: scores should reflect the substance and maturity of a company’s practices, not just the presence of disclosure.
Our ambition is therefore to:
Provide a more meaningful benchmark by creating greater differentiation between levels of AI Governance maturity, enabling investors to identify strengths, gaps, and emerging risks while giving companies richer and more actionable feedback.
Generate deeper insights by translating narrative and qualitative disclosures into structured, comparable data that can support stewardship, engagement, risk assessment, and portfolio-level analysis. This is underpinned by a transparent methodology combining AI-assisted analysis with dedicated human oversight.
Strengthen the engagement process by integrating scoring earlier in the disclosure cycle, enabling companies to receive timely insights, identify potential gaps, and improve the quality of information provided before final assessment.
Align with the evolving responsible business landscape by creating a framework that is interoperable with emerging standards and capable of supporting a more coherent and connected ecosystem of corporate responsibility metrics.
The enhancement lies in the assessment methodology, allowing for more nuanced and decision-useful insights without increasing the reporting burden on participating companies.
The Scoring Framework – 3A Architecture
The scoring methodology is underpinned by a maturity progression, reflecting the idea that meaningful AI governance develops in stages i.e. from recognising a challenge, to establishing a coherent approach to managing it, to demonstrating that this approach has been put into practice and delivered results. This progression is captured through three dimensions: Awareness, Approach, and Action (3A):
Awareness: Does the company recognise and understand the AI governance challenge this module addresses?
Approach: Does the company have coherent and rigorous governance arrangements in place to manage this challenge?
Action: Has the company’s governance been exercised in practice and produced specific outcomes during the reporting period?
Each dimension builds on the one before it, meaning that higher scores are only achievable where a company can evidence not just awareness or intent, but a rigorous approach that has been operationalised and produced tangible outcomes. This structure allows the methodology to differentiate meaningfully between companies at different stages of governance maturity, rather than treating disclosure as a binary exercise.
Not every module is assessed across all three dimensions, for example, some modules apply only Awareness and Approach, or Awareness and Action, depending on which dimensions the module’s evidence can meaningfully support. Where a dimension does not apply to a module, it is excluded from that module’s composite calculation.
Each dimension is scored on an integer scale from 0 to 5, with the number of levels used tailored to how many genuinely distinct maturity states that module’s evidence can support.
Individual dimension scores are aggregated through two levels: Module → Company, producing four company-level scores on a 0–100 scale: an Overall Composite, alongside separate Overall Awareness, Overall Approach, and Overall Action scores.
Example:
Consider a hypothetical mid-cap technology company that has recently begun integrating AI into its core products and established a board-level committee to oversee emerging technology risks. Public disclosures discuss AI-specific risks such as algorithmic bias and model transparency, and senior leaders demonstrate a clear understanding of the opportunities and challenges associated with AI. The company has also introduced governance mechanisms, including an AI ethics policy and formal board oversight, although these structures are still relatively new and not yet fully embedded across the organisation. While the foundations of governance are in place, there is limited evidence that these policies and processes have been systematically tested, audited, or applied to specific AI use cases during the reporting period.
Against this backdrop, the company’s Awareness–Approach–Action profile might score 63 | 58 | 50. This pattern is common among organisations in the early stages of AI governance maturity. Awareness often develops first as companies identify and communicate AI-related risks and opportunities. Governance structures and policies typically follow, resulting in a moderately strong Approach score. However, implementation frequently lags behind, as organisations require time to operationalise new policies, establish controls, and demonstrate outcomes in practice, leading to a comparatively lower Action score.
The differences between the three dimensions provide valuable diagnostic insight. In this example, the five-point gap between Awareness and Approach suggests that the organisation’s understanding of AI-related issues is progressing slightly ahead of the maturity of its governance framework. More notably, the eight-point gap between Approach and Action indicates that while governance mechanisms have been established, there is limited evidence that they have been fully operationalised or embedded into day-to-day decision-making. This type of profile is frequently observed in companies that are building out their AI governance capabilities and are transitioning from policy development to implementation and oversight in practice.
How can you get involved?
We want this methodology to be shaped by the people who use it. There are two ways to contribute during the consultation window:
Join a Consultation Call
We invite our investor signatories to have a conversation with our team to walk you through the framework, the 3A Model, and how the scores and underlying data could support your investment process. Let us know your availability during 24 August – 4 September 2026.
Submit Written Feedback
Prefer to respond in writing? We welcome detailed comments on the draft scoring criteria and scoring framework on this page. Due to capacity reasons, any interested companies can submit their feedback in writing.
We are particularly keen to hear your views on:
- The overall structure of the scoring framework
- Whether the 3A Model effectively distinguishes between varying levels of company performance
- Potential investor use-cases for the scores and underlying data
You can submit call requests or written comments to: aicdi@thomsonreuters.com
Timeline: Feedback gathered during this consultation will directly inform the final framework, which we expect to publish in late September 2026. The public data scorecard will follow in batches through the disclosure window, with the self-disclosure scorecard available early next year.
Resources:
AICDI 2026 Scoring Methodology – Framework and Design Principles
