GeoAI for sustainable supply chains: Insights from industry leaders

Experts from Google, Unilever, and Picterra explore how geospatial AI is accelerating traceability, supporting smallholders, and enabling smarter, data-driven decisions.

As scrutiny over corporate sustainability claims intensifies, so too does the demand for better, faster, and more transparent data. In a recent webinar hosted by Innovation Forum in partnership with Picterra, experts from Google, Unilever, and Picterra explored how Geospatial AI (GeoAI) is transforming supply chain traceability and environmental risk management, particularly in agriculture and forest commodities.

What is GeoAI?

GeoAI combines geospatial data (satellite imagery, aerial photography, drone footage) with machine learning to extract meaningful insights at scale. As explained by Pierrick Poulenas, CEO and co-founder of Picterra, it's about applying machine learning models to extract information from vast amounts of remotely sensed data with the goal of providing "consistent information at scale and with speed."

Yet the challenge is no longer collecting pixels; it's making sense of them quickly and accurately. Most sustainability budgets today are spent on data collection, and GeoAI offers a path to shift that investment toward impact by accelerating data analysis and reducing technical complexity.

Poulenas cautions that GeoAI only works when grounded in context. “You need people on the ground who know the land,” he said. “Otherwise, data remains siloed.” Agronomists and local experts are essential for training accurate machine learning models that reflect real landscapes and supply chain realities.

Picterra’s unique approach places domain expertise at the center of GeoAI. The platform empowers users, such as agronomists or sustainability teams, to develop and implement custom detection models using their on-the-ground knowledge. By keeping humans in the loop, Picterra ensures that geospatial analysis is both context-aware and scalable, without the bottlenecks of external data science teams.

Responsible innovation: How Unilever uses GeoAI

Andrew Wilcox, associate director of sustainability, procurement strategy and insights at Unilever, details the company’s layered approach to AI: combining robust assurance and governance with a culture of innovation. Unilever’s AI initiatives range from global research hubs to company-wide upskilling.

Unilever began working with Google Earth Engine in 2020 to monitor deforestation, helping it to achieve 95% third-party-assured deforestation-free volumes for key agricultural materials. The company also deploys GeoAI as one tool within broader programs that include smallholder support and local partnerships.

Unilever’s current focus includes next-generation applications such as geo-foundation models, retrieval-augmented generation (RAG), and ways to integrate GeoAI outputs into business decisions, without overwhelming users with too much data.

Scaling and democratising GeoAI

Sustainability is part of Google Earth Engine’s DNA, Alicia Sullivan explains, product manager for Earth Engine Sustainability Solutions at Google. Since 2010, it has offered planetary-scale computing power and the world’s largest catalogue of public geospatial data to researchers, governments, and companies.

Sullivan points to efforts to make GeoAI more accessible. Traditionally, geospatial data has been trapped in technical silos. Now, Google is helping partners, including Picterra and others, democratise these tools so even non-experts (such as procurement professionals or smallholder farmers) can extract value. The goal is to scale human expertise, not replace it.

Google is also working with the FAO to support smallholder farmers via tools such as Open Forest Ground, a mobile platform that helps farmers digitize their land data and participate in new markets and compliance environments.

A ground-up approach: the Forest Data Partnership

Both Unilever and Google are founding members of the Forest Data Partnership, which aims to create open, community-driven data layers for key commodities including cocoa, coffee, and palm oil. These tools help companies and smallholders comply with regulations such as the EU Deforestation Regulation by providing shared starting points for mapping and monitoring.

A key principle of the initiative is inclusivity: ensuring data can flow both ways, from satellites down to farmers, and from farmers back up into systems. This bottom-up feedback loop ensures the data reflects reality and empowers local actors to participate in sustainable markets.

Addressing common questions

The session also tackles key questions from the audience (click here to see the questions asked during the webinar):

On carbon offset integrity: Poulenas observes that transparency is essential. “You can’t work with a black box,” he said. Companies need to share their validation methods and data to build trust and prevent double-counting or over-crediting.

On smallholder engagement: GeoAI can be used to pre-map farm plots and share editable versions with local farmers via simple apps, enabling a “give and take” approach rather than one-way data extraction.

On data types: Both Google and Picterra note the importance of blending public and commercial data. Public satellite imagery is essential for broad monitoring, while high-resolution commercial data can be used for targeted analysis when needed.

Key takeaways

GeoAI is powerful, but effectiveness depends on more than just technology. It requires local context, domain expertise, and robust model validation to be effective.
Human insight is essential. From agronomists to smallholders, people on the ground play a crucial role in making AI tools actionable and trustworthy.
Public and open data matter. Democratising access to data is critical for building inclusive, transparent, and effective sustainability systems.
AI is a tool, not the solution. As Wilcox says, the goal is to love the problem, not the solution. GeoAI should support real-world decisions, not distract from them.

While GeoAI offers powerful capabilities for sustainability monitoring, its effectiveness depends on balancing technological innovation with human expertise and local knowledge. For companies pursuing sustainability goals, GeoAI represents not just a compliance tool but a means to better understand and manage complex global supply chains.

The session closes with unanimous agreement: there's still a long road ahead, but the energy and innovation in the GeoAI space is promising and growing.

Want to keep the conversation going with Picterra? Click here to download Picterra’s new whitepaper ‘GeoAI for sustainability. You can also get in touch with the team here.

Click here to view the questions asked during the webinar.

Other useful information:

See Google Earth Engine Initiative to explore Google Cloud Partners and that can help build solutions
Click here to see FAO's digital public infrastructure "WHISP" (What's in that plot?) that supports small and medium organisations to understand deforestation risk.
Click here to see Forest Data Partnership's annual commodity maps on Google Earth Engine

Click to see Alicia Sullivan's talk at Google Cloud NEXT that shows how layers from Forest Data Partnership and others can be used to understand deforestation risk with BigQuery.