Is your institution ready for AI in sustainable finance?
AI in sustainability teams usually starts as isolated pilots that never scale, because no one checked what sits underneath. Score 24 statements across data, analytics, governance and talent in about 15 minutes, and see which use cases you are ready for today.
Twenty-four statements, four layers
Rate each statement from 0 to 3 with the people who know each layer: data, analytics, risk or compliance, and the business users.
Your stack profile, instantly
Your level on each layer, the weakest layer that sets your readiness, and the priority move.
Recommendations by email
The use cases you are ready for and what blocks the rest, an action for every weak statement, a 90-day plan and a recommended starting point.
Which best describes your institution?
Answer for the sustainability and risk work where AI is used or planned, including tools supplied by vendors.
Your answers stay in your browser until you choose to send them.
Is the data AI needs sourced, licensed and traceable?
Score what you can evidence, not what is planned. Where functions disagree, record the lower score: the disagreement is a finding in itself.
The data needed for priority use cases (portfolio, counterparty, hazard, emissions, regulatory text) is inventoried with named owners.
Licences for third-party ESG, climate and nature data permit the intended use, including use in AI models.
External data is matched reliably to your counterparties, issuers or assets (identifiers, locations).
Quality rules for completeness, timeliness and plausibility run on priority datasets, and issues are logged.
Lineage is traceable from source to output, so any number can be explained back to its origin.
Unstructured sources (sustainability reports, transition plans, questionnaires) are stored so they can be searched and cited.
Are methods defined before models, and is output tested?
Each analytical output has a documented methodology, grounded in a recognised standard where one exists, written before model choice.
The choice between rules, statistical models, machine learning or generative AI is justified in writing for each use case.
Performance is measured against defined criteria on held-out, benchmark or expert-labelled data.
Generative AI outputs are grounded in cited sources and tested with evaluation sets for accuracy and fabrication.
Results are reproducible: versions, prompts, configurations and data snapshots are recorded.
Outputs reach users inside the tools and processes where they make decisions, not in a separate dashboard.
Is AI inventoried, classified and reviewed?
Your AI policy explicitly covers sustainability and risk use cases, including tools supplied by vendors.
An inventory records each AI system’s purpose, owner, data, model type and risk class.
Each use case has been classified under the EU AI Act (where relevant) and against FINMA expectations or your supervisor’s equivalent.
Human oversight points are defined: what runs automatically, what is reviewed, what is escalated.
Independent review or effective challenge happens before material use and is documented.
Documentation would satisfy an internal auditor today without reconstruction.
Do you have the people to run, judge and use AI?
A named senior sponsor and a named owner exist for AI in sustainability and risk work.
Sustainability and risk analysts have had AI literacy training suited to the tools they use.
Data and AI specialists understand the sustainability methods behind the outputs they build.
Business users helped design the workflows they now use, and their feedback is acted on.
The capability survives staff turnover and the end of vendor or consultant contracts (documentation, handover, second person).
Roles, skills and training needs for AI are part of workforce planning, not left to individual initiative.
Download your results and recommendations
- Your stack profile: level and next move for each layer
- The use cases you are ready for today, and the layer that blocks each of the others
- An action for every statement below “Established”, and a 90-day plan for your weakest layer
- What applies to your institution type, and a recommended starting point
Thank you. We have your results and will email your report, with a link to pay, within one working day.
Review your results with a senior practitioner
A 30-minute call to compare your results with supervisory expectations. We will tell you which starting point fits, or whether you need outside support at all.
Book a callWhat the Sustainable Intelligence Stack adds to the standards you already follow
- FINMA Guidance 08/2024 sets expectations for governance, inventory, risk classification, data quality, testing, explainability and independent review; the stack spreads these checks across L1–L3 with evidence prompts.
- The EU AI Act (Regulation 2024/1689), as amended by the 2026 Digital Omnibus, sets risk classes, high-risk obligations for stand-alone systems from 2 December 2027 and an AI literacy duty; the stack classifies each use case early (L3) and measures literacy as a layer (L4).
- NIST AI RMF 1.0 and ISO/IEC 42001 define AI risk management and management systems; the stack sizes controls to sustainability use cases.
- Model risk management (for example SR 11-7 and PRA SS1/23) and sustainable finance methods (PCAF, NGFS, TNFD, ISSB) set validation and meaning; the stack documents methods before models and judges analytics against the domain method.
Edition November 2026. Self-assessed results are general guidance, not regulatory, legal or investment advice. Regulatory references reflect the status known to Artamis in November 2026; confirm current requirements for your jurisdiction and entities. This is a self-evaluation, not an audit, a validation or a legal opinion on EU AI Act classification. Minimum levels in the use-case library are Artamis guidance; your context may differ. Artamis does not sell an AI software platform.
Prefer to talk it through first?
Bring your results, or your questions, to a 30-minute conversation. We will tell you where to start.