Ambi Robotics Inc. said its AI agents solved in 10 hours a package-placement problem that would have taken engineers several weeks to address. It did this using a new Graph-as-Policy-based Agentic Robotics harness within AmbiOS.
The Berkeley, Calif.-based company said it has deployed the system across 30% of its U.S. fleet. This includes at the operations of a Fortune 50 package-shipping business.
Improving performance on commercially-deployed robots requires significant human effort, noted Ambi Robotics. Engineers must review key data offline, including videos, images, sensor data, and logs, hypothesize root causes, and test updates. This process is slow and requires significant domain knowledge.
AmbiOS Agentic Robotic Harness takes a new approach
“Agentic robotics” offers a new approach, said Ambi Robotics. The idea is to use AI coding agents to automatically analyze data, identify problems, and solve them to improve robot performance.
This approach combines rapid advances in AI coding with classical model-based robotics modules and learned model-free policies. It requires agentic robotics “harnesses” to guide AI coding systems to solve the real robotics problems, according to Ambi.
Several new research papers, including Code as Policy (CaP) and Graph as Policy (GaP), show promise for rapidly generating reliable robot control systems with agentic harnesses that constrain agents to use libraries of relevant robot skills.
Building on these ideas, Ambi created the AmbiOS Agentic Robotics Harness. It builds on GaP to update computation graphs that are composed of existing robot skills.
The harness gives agents access to two key pieces of infrastructure. First, agents access datasets of real robot events from the cloud. Agents can filter these by a set of known success and failure modes.
Second, agents use a simulation environment to replay these events with new self-generated behaviors and test graph edits, evaluating the outcomes in the simulation. The harness constrains how agents work, ensuring that the final result can be executed without compromising safety checks and error handling.
Ambi Robotics AI agents address a package problem
The AmbiSort robot picks, scans, and places random packages with a robot arm. A gantry receives the package and sorts it into a bag based on data like the ZIP code. The target bags have a fixed opening size through which items must fit to fall into the bag without jamming.
Ambi applied the AmbiOS Agentic Robotics Harness to a problem where subtle edge cases affect critical operational metrics: throughput, sort accuracy, and uptime. Millimeter-level inaccuracies can have a significant effect on throughput, the company noted.
Packages that are on the boundary of fitting into the bag opening are challenging because the robot arm cannot estimate item dimensions and pose with perfect precision due to sensor noise and item deformation.
Furthermore, the robot prefers placing items with the short side going into the bag to maximize space utilization. As a result, these edge case items are too wide to fit through the bag opening, forcing the system to reject and retry packages, losing valuable time.
Ambi estimated that this issue was reducing capacity by approximately 31,000 sorts per year per robot for a typical facility running 12 hours a day, six days a week. Fixing this issue with hand-tuning would take several weeks of engineering time.
The company used the AmbiOS Agentic Robotics Harness with Anthropic’s Claude models including both Sonnet 4.6 and Opus 4.8. The agents analyzed images and data from 500 events over the previous week of production. Half of these were the known failure mode and 50% were selected at random to avoid regressions. Within 10 hours, the agents generated and evaluated three solution hypotheses.
Inside Ambi’s results
Ambi deployed these three hypotheses into real production A/B tests in customer warehouses. Each solution was compared to a baseline software version. Here are the results after 10 days:
| Hypothesis | Estimated Change in Sorts per Year per Robot |
| H1 | −374 |
| H2 | −1,497 |
| H3 | +15,725 |
Table 1. Comparison of throughput and estimated packages per year across three A/B test solutions deployed to real production robots.
For H3, the agent noticed that for large, dangling bags and paper mailers, the placement plan needs to carefully trade off orienting the item to fit in the destination bag and avoiding placements where the item could get stuck in grates on the placement platform.
It optimized this tradeoff and decided to route more packages through a subgraph for planning dangling item placements because it results in a more predictable final orientation. In addition, the agent changed the flat item subgraph to only optimize for the final orientation of the item. This is because it is easier to predict the final resting position of flat items than for bags.
In real production A/B tests, H3 increased throughput by 4.2 packages per hour. Over 12 hours per day and six days a week (standard operation), this corresponds to more than 15,000 packages per year per robot.