The future is now: AI for nature analysis and action :: Environmental Finance

By Thomas Moran, Head of Nature and Biodiversity Products, and Vineet Gupta, Chief of Product and Technology at GIST Impact

From overwhelmed to empowered

From global banks to boutique investment firms, we hear the same refrain from financial professionals: “We need to use more nature data, but where do we even begin with all this information?” Most financial institutions have processes in place for climate assessment, but the transition to incorporate nature and biodiversity feels daunting—as if it requires a whole new set of expertise they don’t currently possess.

We see AI as having tremendous potential to contextualise and simplify this topic for sustainability professionals, analysts, and their stakeholders. This isn’t a future promise; AI applications are already transforming how professionals work and make decisions with complex environmental data.

Today: AI as assistant

Current AI applications primarily serve as assistive technologies, with significant human involvement still required for the core of work at financial institutions: regulation, assurance, and risk management. Two key applications demonstrate this value for financial institutions:

1. Making regulatory complexity manageable

One of our first customer-facing applications of AI is for CSRD compliance. We maintain a database of over 17,000 companies and generate CSRD double materiality ratings using a combination of company disclosures, public data, and third-party business intelligence. This approach is particularly valuable for challenging sections like ESRS E4 (“Biodiversity and Ecosystems”), where information is scarce or spread across many data sources.

We apply AI to scan massive sources of company information, materiality guidance, web searches, and other sources, then integrate this into an evaluation framework. This enables us to not only provide high quality materiality scores at the sub-sub-topic level, but to also explain the rationale, summarise findings in a consumable way, and to provide references. Our customers and stakeholders have responded positively to these results as they not only serve the ultimate reporting goal, but also create a comprehensive and traceable source of information.

A powerful aspect of this capability is adaptability. When guidance or regulations change, as they did with CSRD recently, adapting to new requirements is a straightforward matter of updating the AI process rather than manually redoing an immense amount of analysis. While fully autonomous AI-generated content isn’t yet reality, we bridge this gap with a team of highly specialised analysts—experts in environmental and scientific fields—who provide sector-specific knowledge and contextual understanding to ensure accurate, meaningful insights.

2. Bringing asset-level analysis to scale

There is widespread agreement that nature risks and impacts need to be assessed down to the level of individual physical assets and operations. Historically, asset data was difficult to obtain and even harder to analyse reliably. While anyone can examine satellite images of a single location today, making sense of what’s happening at millions of asset locations across a global portfolio presents a significantly scaled-up challenge.

This challenge is becoming more and more solvable with foundation models—AI systems, often open source, that can be trained with commonly available satellite data to recognise different types of land use or buildings based on a few examples. The power of this approach is that organisations can perform sophisticated remote sensing analysis—such as quickly examining millions of assets to identify locations potentially associated with air pollutants—without requiring a dedicated data science team.

2-3 Years: From Predetermined to On-Demand Analytics

As AI capabilities advance, capabilities will shift from predetermined analyses to on-demand intelligence.

Portfolio analytics at your command

Our CSRD use-case demonstrates how AI can extract and analyse predetermined types of information from diverse sources. This works well when many users have similar needs, but the transformative capability comes when analysts can request arbitrary, highly specific information and analysis on demand.

Consider this scenario: a major agricultural country moves to embrace GMO crops, and an analyst needs to assess transition risks by identifying which food processing companies with supply chain links to that country also maintain policies against GMO use. As we enhance our ability to retrieve and analyse multimodal data, fewer limitations will exist on the types of questions that can be asked and immediately answered.

Enhanced supply chain visibility

Beyond asset-level data, supply chain information represents another major challenge for nature sustainability analysis. While portfolio-level supply chains with investor-grade accuracy remain a future goal, we expect rapid progress in piecing together the main components of global corporate value chains.

The first step is moving beyond high-level input-output models like EXIOBASE, which provide general insights but weren’t designed to model individual company value chains. Today, we can assemble the main components and geographies of a business’s value chain through disclosures and third-party intelligence. By automating this process across tens of thousands of businesses and analysing cross-sector connections, we’ll develop a more nuanced and accurate view of value chains.

5+ Years: From Measurement to Action

The longer-term vision involves a shift from measurement to action.

Currently, the focus remains on measurement and understanding. The future will center on decisions, actions and delivering results. We anticipate AI driving intentional decisions related to biodiversity-linked supply chains, exposure detection in real-time, and proactive risk management.

We’ll move beyond reporting and disclosure toward action—making proactive decisions on financial risk reduction, decarbonisation, supply chain optimisation, and building resilient agreements. Impact measurement, data collection, and scoring will eventually be fully automated.

As capabilities evolve, we’ll approach real-time analysis of changing conditions, enabling rapid responses to events like wildfires, political shifts, or new climate models. This will create real-time disclosure capabilities and dashboards supporting swift decision-making for corporates and investors. The implications of better information will reverberate through value chains, benefiting both business and nature.

Navigating the AI-enabled future

Many sustainability professionals and data providers still discuss AI in future terms—as something on the horizon or not yet ready for prime time. We hope to have clarified that valuable applications already do exist, and are expanding at a rapid pace, especially for combining and synthesising information. And as the saying goes—when it comes to AI—the best is yet to come.

We expect the nature and balance of AI-human collaboration to evolve: from AI assistance today, to collaborative systems next, to more autonomous capabilities in the future.

When engaging with data providers, don’t hesitate to ask for what you really want rather than limiting yourself to what seems possible, because what is possible is changing faster than ever before.