Graft Systems

American wineries lose millions each harvest season to guesswork. Winemakers walk rows of vines and estimate by eye how much fruit they'll bring in, a skill built over decades that retires when the winemaker does. Graft Systems is a team with firsthand experience on harvest floors from Bordeaux to Napa, building an AI-powered tool that estimates total grape cluster volume from footage. Our SaaS platform delivers precision yield data to any winery, from a two-person family operation to a 500-acre estate, at a fraction of the cost of enterprise agtech. We are building the future of harvest intelligence.

The US wine industry comprises more than 11,000 wineries, generating over $90 billion in annual revenue, and the global market is on track to double from $550 billion to $1 trillion by 2033. Yet behind every bottle is a harvest, and behind every harvest is a yield estimate still made by a winemaker walking vine rows and guessing by eye. We size our immediate US addressable market at the intersection of 11,000 wineries within a $1.2 billion precision viticulture software segment growing at 13.9% annually. Our beachhead: 500 small and mid-sized NorCal wineries that cannot afford enterprise agtech but desperately need real data. Graft Systems changes how that estimate gets made. A vineyard worker records grapevines with a camera attached to their already operating machinery; our model returns an instant volume estimate that accounts for leaf occlusion and varietal-specific cluster morphology. The core insight: machine learning can distinguish what the human eye cannot: the density differences between Dijon 667 and 777 Pinot Noir clones, the clusters hidden behind a dense canopy, the variance that accumulates across a 20-acre block. We are already within 20% accuracy pre-ground-truth and have a standing invitation from William Selyem to validate our model in the field this harvest season. Our direct competitors, Cropty and generalist platforms like Trimble Agriculture, are designed for corn, soybeans, and apples. They fail with wine grapes because they cannot account for irregular cluster morphology, vertical hang patterns, or vine training systems such as VSP or Cordon Training. Our moat is data specificity: every scan run through Graft becomes a labeled data point that sharpens our model. This cycle, more users, more data, better accuracy, compounds in a way that a generalist competitor cannot replicate without years of grape-specific data collection; they have no incentive to pursue. Our go-to-market path runs through existing relationships. We are already in conversation with heads of winemaking at Kendall Jackson Estates, William Selyem, Honig Winery, and Monroy Wines, collectively producing over 75 million bottles annually. These anchor partners validate accuracy benchmarks, generate proprietary training data, and serve as reference customers for regional expansion. We would sell on a tiered SaaS model, lower-cost entry plans for smaller operations, and scaling for bigger estates. Bart Haycraft, Head Vineyard Manager at Kendall Jackson, set our bar directly: "If you make it within 5% accuracy, you are in business." We plan to be there by harvest 2026. Benson brings firsthand wine industry experience, having worked harvest floors at Troplong Mondot in Bordeaux and staffed the Nantucket Wine Festival, and leads product direction and winery relations alongside our technical roadmap. Viraaj is our AI engineer, responsible for building and iterating the computer vision model at the core of our product. Kayan built and manages our industry relationships, having sat across from the heads of winemaking at Kendall Jackson, William Selyem, Honig, and Monroy, and leads the business and commercialization strategy. We have not just studied this problem; we have been in the rooms where it lives. No generalist is coming.