Harvest AI Talk to us

Artificial intelligence for agribusiness

Every field plot, read before the harvest.

Harvest AI combines satellite imagery, weather and farm management data to forecast the yield of every crop, approve rural credit with better information and prove grain origin to the markets that pay for it.

Explore the platform Investment thesis

Launch crops: soy and corn · Starting in Brazil

Plot
T01
Crop
Soy
Area
0 ha
NDVI
0.00
Est. yield
0 bags/ha
Status
On track
Simulated farm · illustrative dataHover or tap a plot
R$ 3.2 tn

Brazilian agribusiness GDP in 2025, up 12.2% from the previous year.

CNA and Cepea
R$ 525 bn

2026/27 Plano Safra for commercial agriculture, the rural credit Harvest AI helps underwrite.

Canal Rural
350 Mt+

Grain output in the 2025/26 crop year, a record according to Conab.

CNN Brasil, Conab data
5.07 m

Agricultural establishments in the country, each with plots still managed with little data.

IBGE, 2017 Agricultural Census

The problem

Farms generate data. Decisions still arrive late.

Three decisions that move billions every season still rest on sampling, paperwork and site visits.

Forecasting

Yield estimated in the dark

Estimates depend on field sampling, spreadsheets and the experience of whoever visits the crop. The error only shows at harvest, when there is nothing left to adjust.

Credit

Slow, expensive underwriting

Banks and cooperatives assess risk with documents and visits, with little information about the crop itself. The farmer waits, and risk is priced for the worst case.

Origin

Origin is hard to prove

Buyer markets such as the European Union require traceability and no deforestation. Today the proof is a pile of documents scattered across farm, warehouse and trading house.

Platform

One data layer, four products.

Everything starts from the same living map of each plot. Each module turns that map into a different decision for a different customer.

Crop monitoring, plot by plot

Tracks vegetation vigor through the whole cycle and compares each area with its own history and with neighboring plots. When something drifts from expectations, the alert arrives with the exact location.

What it delivers

  • Vegetation indices such as NDVI, refreshed with every satellite pass
  • Alerts for vigor drops, water stress and planting gaps
  • Comparison with the historical average and with similar plots

Who uses it

  • Farmers, agronomists and advisory firms
Vegetation vigor curve across the season, with a vigor drop detected in plot 07 0.80.60.40.2 OctNovDecJanFebMarApr today Vigor drop plot T07 current season historical average

Illustrative example · NDVI of one plot

How it works

From satellite to decision in three steps.

The process is the same for every module. What changes is the question the customer asks of the map.

1

Ingestion

Public imagery such as Sentinel-2, at 10-meter resolution, plus weather series, soil data and the history supplied by the farmer flow into a single database organized by plot.

2

Modeling

Time-series and computer vision models, calibrated by crop and by region, turn the raw signal into vigor, crop stage and expected yield.

3

Decision

The result arrives as a phone alert, a credit opinion for the bank, a dossier for the exporter or an API call inside the system the customer already uses.

Data sources

SatelliteWeatherSoilFarm managementClimate zoning (ZARC)Rural registry (CAR)

Market and thesis

Why now, and why Brazil.

Scale, capital and regulatory pressure meet at a point where crop data stops being a differentiator and becomes a requirement.

Scale that is hard to match

Brazil is the world's largest soybean producer and exporter. Every point of accuracy in forecasting and in risk is worth a lot of financial volume.

Capital that still decides on paper

Hundreds of billions of reais in rural credit are granted every year through largely manual underwriting. Automating the agronomic opinion saves time and lowers the cost of risk.

Traceability became market access

International buyers demand proof of origin. Whoever delivers the dossier keeps the sale, and whoever cannot loses the contract.

Open data and cheap compute

Public satellites and the cloud made it feasible to monitor millions of hectares at low marginal cost. The edge is in the models and the accumulated data, not in capturing the image.

Planned revenue model

Farmers and cooperativesSubscription per monitored hectareMonitor · Forecast
Banks, fintechs and agri fundsFee per opinion and per monitored loanCredit
Trading houses and exportersAnnual license by traced volumeTrace

Competitive advantage

Each season's data makes the next model better.

History accumulated by plot

Every season adds a layer that a new entrant does not have and cannot buy.

Regional models

Yield and weather change from one municipality to the next. Local models beat generic ones.

One dataset, three payers

The same map earns revenue from the farmer, the financial institution and the exporter.

Distribution through existing relationships

Cooperatives, banks and trading houses bring the platform to the farmer, with no farm-by-farm selling.

Reinforcing loop: more plots produce more accurate models, which produce better decisions, which attract more partners and lead to more plots More plots More accuratemodels Betterdecisions More partners Data per plot

The platform's reinforcing loop

Roadmap

Four phases, one validation at a time.

Each phase only moves forward when the previous one proves a result measured against the real harvest.

  1. Phase 1 · 0 to 6 months

    Foundation

    • Monitor and Forecast for soy and corn
    • Pilots with farmers and one cooperative
    • Forecast error measured against the harvest
  2. Phase 2 · 6 to 12 months

    Credit

    • Harvest Credit with a partner bank or credit cooperative
    • Integration with ZARC and CAR
    • API for partners
  3. Phase 3 · 12 to 24 months

    Origin

    • Harvest Trace with trading houses and exporters
    • New crops: cotton, coffee and sugarcane
  4. Phase 4 · 24 months onward

    Expansion

    • Latin America
    • Parametric insurance products with partner insurers

Initial plan, subject to validation with customers and partners.

Risks

What has to go right, and how we handle it.

Model accuracy

Every pilot publishes its forecast error against the real harvest. Expansion into new regions and crops depends on validated metrics, not on promises.

Adoption in the field

Entering through intermediaries such as cooperatives, banks and trading houses removes the friction of selling farm by farm and shortens the sales cycle.

Data and regulation

Compliance with Brazil's data protection law (LGPD) and with Central Bank rules for rural credit from the first design, with farmers in control of who can access their farm data.

Contact

Let's build the next season of data.

We are opening conversations with investors, cooperatives, banks and farmers who want to take part in the first pilots.

fred@ledgernest.com.br
InvestorsDeck, financial model and pilot plan.
Cooperatives and banksForecasting and credit analysis pilot.
FarmersJoin the waitlist for the first monitored plots.