A standard Custeio credit request at a Banco do Brasil rural lending branch in Varginha or Pouso Alegre passes through at least seven documentation steps before a decision. Assembling the file, usually including the INCRA land registry, current CAR status, three years of DAF income records, two years of bank statements, and a SERASA clearance report, takes three to six weeks for a full-time grower who cannot spend several days traveling to municipal offices and bank branches during planting season.
Culttivo delivers a credit decision within 48 hours. The first question from growers and agricultural finance professionals is consistent: how can 48 hours provide a rigorous review? It is a reasonable concern. In financial services, faster decisions can sometimes reduce underwriting quality. For us, the 48-hour schedule comes from automated data handling, not less diligence. This is how the pipeline works.
Why Traditional Agricultural Credit Takes Time
To assess whether 48 hours is realistic, it helps to identify why conventional agricultural credit moves slowly. Four separate stages consume most of the time.
Stage one is borrower-led document collection. Obtaining INCRA registration, current CAR status, income records, and bureau clearances cannot usually be done in one day by a grower working full-time on the farm, especially when the municipality is 40 to 90 minutes from the closest bank branch. This step alone often requires two to three weeks.
Stage two is the credit analyst's document check. Each item must be matched to bank-specific checklists, missing information must be identified, and extra documents must be requested. At a branch handling a queue of applications with a small team, the review may sit for several days before starting and take several more days to finish.
Stage three is committee consideration for loans above the branch approval limit. Many rural credit requests need signoff from a regional credit committee that meets on a set schedule. That can add a week or more, regardless of the application's quality.
Stage four covers notarial and legal work: preparing the loan agreement, signing it, notarizing it where needed, and authorizing disbursement. Even after approval, this final step generally takes three to five business days.
These stages do not materially improve the risk assessment. Risk is shaped by the grower's production capacity and cash flow pattern, both of which can be evaluated through agronomic data. Documentation provides verification and compliance around that assessment, and those are the activities automation can replace.
Layer 1: Data Ingestion Instead of the Document Stack
After a grower applies through our portal, the pipeline's first layer starts on its own. No person takes part in this step.
The grower marks the plot boundary in the portal with the mobile GPS tool or uploads a KML file from an existing farm registry. After submission, the system checks the NASA SRTM elevation model for mean plot altitude, checks the EMBRAPA national soil classification database for the soil beneath the polygon, and begins retrieving Sentinel-2 NDVI imagery for the location covering the prior six months. The NDVI retrieval uses ESA's Copernicus Open Access Hub API, with 10-metre multispectral imagery refreshed on a five-day revisit cycle. We download and process the latest cloud-free passes available in that six-month period.
At the same time, the application records the grower's reported harvest information: delivery quantity by season, in 60-kilogram bags or per hectare, the cooperative or buyer that received the coffee, and any certified quality grade details. If the grower belongs to a cooperative with an authorized data-sharing agreement, we request delivery records from the cooperative's system.
This layer finishes within 30 to 90 minutes after submission, based on NDVI availability and the cooperative's response time. The system saves every item as a structured input record for scoring.
Layer 2: The Scoring Engine
The scoring engine uses the completed input record to calculate production risk on a 1-to-100 scale. The model was trained with Sul de Minas altitude coffee production data from regional cooperatives and calibrated using observed yield results from our early-access pilot cohort.
The calculation itself is quick, with the model running in seconds. Computation does not limit the 48-hour period. Instead, that window gives a human reviewer time to inspect the application and flag inputs needing follow-up before the decision.
Manual review flags can be raised by NDVI patterns showing an unusual event in the prior season, possibly frost damage, pest pressure, or replanting that satellite data cannot separate without more context; an altitude-to-yield relationship outside the expected Sul de Minas range; or a yield record containing fewer than two complete seasons of verified data. When a flag appears, someone on our credit team reviews the relevant inputs and contacts the grower if more information is necessary.
Applications scoring above 65 without manual review flags move through automatically. Applications scoring from 40 to 65, as well as flagged applications, receive a short manual review within the 48-hour period. Applications below 40 are declined automatically and the grower receives a written explanation.
Layer 3: Decision Results and Offer Terms
Once scoring and review are complete, the grower receives a structured credit offer through the portal and by email. It specifies the approved loan amount, monthly interest rate, based on the farm-size pricing tier, the loan term in months, and a full repayment schedule with each payment amount and due date matched to the expected harvest cycle.
The offer remains valid for 30 days. The grower accepts digitally by signing the terms agreement in the portal with a digital signature compliant with Brazilian electronic commerce law. Within five business days of acceptance, funds are sent by PIX to the grower's registered bank account or cooperative account.
From a completed application to available funds, the usual total is 5 to 9 business days: 48 hours for the credit decision and 5 business days for processing the disbursement after acceptance. A grower seeking working capital for input purchases in February or March can therefore apply in the first week of the month and receive funds before input season reaches its peak.
What a 48-Hour Decision Does Not Cover
We also need to define what we are not saying. A 48-hour credit decision using satellite and soil data does not replace every possible form of due diligence for a rural credit application. Some genuine risk factors remain outside the pipeline's full view.
We do not independently confirm land tenure. Growers declare their land title status, and we search available public databases for registered encumbrances. However, a complicated land matter absent from those databases would not be detected by the pipeline. A grower involved in a valid land dispute with an unregistered third-party claim presents collateral risk that we cannot fully evaluate through software.
We do not confirm every income source. A grower with significant off-farm income that changes debt service capacity, or significant off-farm debt that changes the overall credit picture, may not disclose it fully in the application. We ask about these items, but for applications below our Producer tier, we do not verify them independently.
We do not measure market price risk. If a grower committed the harvest in advance at a low fixed price, removing the price upside that would ordinarily support repayment, production data would not reveal that particular risk. The application asks about forward sale commitments, but the answer depends on self-disclosure.
Because of these limits, the model is calibrated to the risks it can observe well and takes a conservative position where visibility is weaker. Scores from 40 to 65 lead to lower loan amounts partly because that band signals greater model uncertainty, not just greater production risk. The 48-hour timeline results from processing observable information efficiently. The rigor comes from acknowledging what cannot be observed and setting terms accordingly.