How Natural Intelligence Empowers Farmers in Emerging Economies

When Cerulean invested in Epoch, we saw a company tackling one of the most important blind spots in global climate and supply chain infrastructure: the first mile.

The first mile is where commodities are grown, harvested, processed, and first enter global supply chains. It is also where many of the hardest sustainability questions begin. Which farmers or facilities are in a buyer’s sourcing area? Which plots are compliant with deforestation rules? Which are organic or regenerative or decarbonizing? Where are emissions, water risk, biodiversity loss, or land-use change concentrated? For global food, agriculture, and forestry supply chains, these questions are not peripheral. They are increasingly core operating, regulatory, and financial questions.

Agriculture and forestry-linked supply chains contribute over 9 Gt of CO2e emissions a year, with over 6 Gt of those emissions linked to commodities produced in the Global South – part of the world that is marred by measurement and data challenges.

This week, Google Earth and Earth Engine highlighted Epoch’s work applying AlphaEarth Foundations satellite embeddings to agricultural facility detection. The case study, written by Epoch founding engineer Jake Wilkins, shows why we continue to believe Epoch is building a category-defining data layer for the first mile. Starting with just 50 labeled examples, a laptop, and Google’s AlphaEarth Foundations Satellite Embedding dataset, Epoch detected nearly 70% of Sumatra’s known palm oil mills in a single afternoon.

That result matters because first-mile mapping has historically been slow, manual, and expensive. Large, verified facility datasets exist for a few high-profile commodities and regions, but most global commodity supply chains still lack accurate, continuously updated, plot- and facility-level visibility. Epoch’s work points toward a different operating model: use AI and geospatial embeddings to bootstrap country-scale intelligence quickly, then apply expert validation, post-processing, and workflow integration to make the data useful for compliance and commercial decisions.

The technical insight is powerful. AlphaEarth Foundations compresses multi-source Earth observation data into 10-meter, 64-dimensional embeddings designed for classification, clustering, and change detection. Google DeepMind describes the model as integrating Earth observation data into a unified digital representation, and Google’s Earth Engine catalog describes the dataset as annual, global, analysis-ready geospatial embeddings.

Epoch used those embeddings to search for the satellite signature of palm oil mill effluent lagoons — a distinctive feature associated with palm oil processing infrastructure. After two training iterations and simple spatial clustering, the team reduced a country-scale search problem to a few thousand candidate sites for review, while recovering roughly 70% of documented mills in Sumatra.

This is exactly the kind of leverage Cerulean looks for: AI applied to data-rich but under-digitized markets where better information can change capital allocation, compliance, and incentives. Cerulean’s “Natural Intelligence” investment thesis focuses on founders using AI, proprietary data, compute, and fintech to power the sustainable transformation of large markets that have been resistant to digital transformation — including soft tropical commodities, biodiversity, and natural capital.

Epoch’s platform is already positioned at the intersection of these forces. The company describes its product as a first-mile data layer for supply chains: one API integration for environmental and compliance use cases, without additional supplier questionnaires. Epoch monitors more than 15 million farms and plantations, analyzes more than 250 million hectares, and supports more than 24,000 corporates on a path to sustainability through platform partners and supplier workflows.

The market pull is also intensifying. EUDR and related regulations require that companies understand where their commodities originate and whether they are linked to deforestation. Moreover, companies subject to EUDR must prove that certain commodities and derived products are not sourced from land deforested after December 2020, and that compliance requires geospatial coordinates for sourcing areas. Google highlighted Epoch as one of the specialized providers using Google Cloud, Forest Data Partnership datasets, and Earth Engine to detect target commodity plots and automatically delineate plots where supplier-provided data is missing.

For Cerulean, the broader significance is not only compliance. Compliance may be the forcing function, but the long-term opportunity is larger: a trusted first-mile data infrastructure layer that allows markets to value better sourcing, lower risk, more resilient production, and nature-positive outcomes. that means valuing regenerative agriculture at scale and incentivizing decarbonization and climate positive land stewardship globally.

When first-mile data becomes faster, cheaper, more transparent, and easier to integrate, companies can move from broad sustainability claims to targeted action. They can identify where risk is concentrated. They can focus remediation and incentives where they matter most. They can build audit trails. They can connect capital to measurable environmental outcomes. And they can begin to treat nature, land use, and supplier resilience as core enterprise data — not as a periodic reporting exercise.

whether this natural intelligence is used to incentivize or reward regenerative agriculture, Circular material practices, Renewable energy deployment at the edge of the global economy, The critical first principle is to avoid creating new work for land stewards and commodity producers while guaranteeing more stable and higher incomes.

That is why Epoch’s Google Earth/Earth Engine feature is more than an impressive technical case study. It is a glimpse of what AI-native climate infrastructure can look like: fast to scale, rigorous verification, and detailed down to the individual farmer contributing to sustainable supply chains.

We are proud to back Epoch as they build the first-mile intelligence layer for more transparent, resilient, and sustainable global supply chains.