$2.1M to finish an AI grading system for use in the field — a Phase III commercialization award (USDA)

1What "grading" is doing

The United States has long operated systems that assign grades to agricultural products and meat. Behind the labels a shopper sees lies the work of applying defined criteria, performed by people. That judgment takes expertise and has carried persistent problems: variation between graders and the difficulty of staffing the work at all.

The motive for bringing in automation built on image recognition is clear — holding the application of criteria steady, and raising throughput.

2Phase III is where research moves to the field

Phase III in the scope names a stage in the U.S. small-business research and development framework. Phase I establishes feasibility, Phase II builds and demonstrates a prototype, and Phase III moves to commercialization and fielding.

The work this contract lists — model development, software integration, hardware configuration, performance validation, deployment — is precisely the process of turning something that works into something usable on site. That this stage arrives as a procurement contract rather than research funding is what characterizes Phase III.

3The condition "fully functional, field ready"

The scope states its completion condition as fully functional and field ready. Achieving accuracy in a lab is not the same as running stably under the lighting, speed, soil, and vibration of a working facility. Performance validation and hardware configuration appear in one contract because closing that gap is the work. The specific commodities and the criteria being judged cannot be confirmed, since the record is truncated.

4The fiscal years the records fall in

Sorting the 222 federal spending records this site holds as of 2026-09-01 by the fiscal year in which the contract began gives FY2024 the most at 54, then FY2025 at 43 and FY2026 at 36. Further back, FY2020 has 27 and FY2023 has 26.

More than half fall in the last three years. That points to procurement touching AI having grown recently, and equally to this site's collection being thickest over that period. The run of counts cannot be read straight off as the movement of a market.