On robots and platforms
Why should you leave execution data?
More important than a good proposal is checking what was actually implemented and what results and exceptions were left behind.
Robots and learning/work platforms are very different on the surface. However, for a product to create value in the field, the user's intention, the system's judgment, and actual actions and results must be connected. The material that confirms this connection is execution data.
A lot of logs are not execution data.
Storing a lot of server logs or sensor data can only increase costs and privacy risks. Minimum data should be designed around the target action, success/failure conditions, exceptions to be checked by humans, and next decisions.
Robots need both state and context
Position and speed alone cannot determine the reason for stopping. Mission, route, obstacles, safety events, battery, communication status and operator intervention must be linked together to analyze replanning failures or bottlenecks.
Usage flow is important for platforms
Starting learning, completing assessments, non-responses, checking textual evidence, and next actions are more important than page views. We need to see different results: learning flow for PassAI, training operation for WorkCheck, exploration of evidence for Desk, and question-based activity for Kids.
Records human judgment
It is necessary to distinguish whether the system implemented the suggestions as is or whether the user modified or rejected them. By treating human verification as a critical process for safety and quality rather than a failure, you can avoid over-promising automation.
Improveable data structures
- Goal: Decide which issues and results you want to check.
- Action: It records what the system and users actually did.
- Result: Measure completion, delays, failures and quality.
- Context: Connect environment, version, target and constraints.
- Review: A person checks and decides on the next change.
Why is it important for commercialization?
A record showing what has been improved and how far it can be expanded is more reliable than a single statement that the PoC was successful. For services in operation, you can determine the next product and scope of support by understanding the reasons for repeated use, departures, and exceptions.
Privacy and Security
Collect only the data you need and be clear about its purpose, retention period, and access rights. Sensitive information such as email, inquiry content, or child data is not included in analytics events, and public sites do not load external analytics scripts without a measurement ID.
RODIX AX connection
AX, which RODIX refers to, is a structure where technology does not end with creating an answer, but leads to execution and results confirmed by humans. The different businesses of robotics and platforms are connected by the common principle of action data.
Note
We referred to the RODIX final work instructions, platform-specific operational data, and robotics PoC/verification documents.