Project Update 1
Piping UT Demonstration Project Launch
ControlPointAI launches a fictionalized submarine piping UT demonstration to test AI-assisted inspection analysis, human technical authority, runtime revalidation, and traceable engineering decision-making.
By Wayne Couch ·
Controlled Baseline — Active Demonstration Project 28 August 2026
Project Overview
ControlPointAI is launching a formal demonstration project built around a fictionalized submarine piping ultrasonic-testing (UT) inspection and disposition workflow.
The project will demonstrate how AI can assist with receiving, organizing, evaluating, and presenting inspection evidence within a controlled engineering process while preserving traceability between source evidence, AI-generated interpretation, recommendation, human technical authority, disposition, and the resulting controlled record.
What the Project Is Testing
The demonstration will evaluate how AI can support an engineering workflow by:
- Receiving and organizing synthetic UT inspection evidence and approved requirement/context data.
- Checking the evidence package for completeness, consistency, provenance, and missing information.
- Distinguishing observed evidence from inferred state, analytical interpretation, recommendation, and authorized action.
- Supporting comparison or trending of synthetic inspection data where permitted by the demonstration baseline.
- Identifying conditions that warrant human review, additional evidence, escalation, or revalidation.
- Preparing a draft engineering disposition or UT work product from the controlled evidence package.
- Maintaining traceability between evidence, AI output, human decision, authority basis, and the final controlled record.
AI Authority Boundary
Within this demonstration, AI may assist, analyze, flag, organize, compare, recommend, and draft.
AI will not independently:
- Declare a component acceptable for continued service.
- Authorize repair, replacement, additional inspection, or return to service.
- Waive, alter, or reinterpret an applicable technical requirement as an authoritative act.
- Create missing inspection evidence or treat inference as observed evidence.
- Approve its own generated engineering disposition or UT work product.
- Exercise final technical, certification, release, or execution authority.
Human Technical Authority
Final engineering judgment remains human.
A designated human technical authority will review the evidence package and AI-generated material, determine whether the evidence is adequate, approve or reject the proposed disposition, direct additional evidence or analysis when required, and authorize the final technical outcome represented in the demonstration.
The project is specifically intended to preserve the distinction between an AI recommendation and an authorized engineering decision.
A Central Demonstration Question
Can the system recognize when new UT evidence materially changes the basis of an existing authorization, stop the workflow before consequential execution continues, and require meaningful human revalidation before expanded or otherwise unauthorized action can occur?
This moves the demonstration beyond a simple “human-in-the-loop” approval model.
The objective is not merely to preserve a human signature or approval click, but to preserve meaningful human technical judgment at consequential control points.
Public-Source / Synthetic Demonstration
The submarine-maintenance context is used to provide a realistic engineering setting. The technical case itself will be built from approved public-source material and purpose-built synthetic data.
The demonstration will not reproduce classified information, Controlled Unclassified Information (CUI), NOFORN material, proprietary information, ship-specific configuration information, actual inspection records, non-public acceptance criteria, controlled procedures, internal work packages, or other restricted U.S. Navy technical data.
Public Navy, DoD, commercial, industrial, academic, and regulatory sources may be used to establish terminology, program context, high-level maintenance and NDT concepts, publicly identified standards, and other information already approved for public release.
Case-specific inspection values, component details, thresholds, decision logic, authority assignments, and resulting demonstration records will remain synthetic unless explicitly identified as approved public-source material.
What Comes Next
The project will now move from the established objective and technical baseline into development of the demonstration architecture and test scenarios.
Planned work includes:
- Mapping the inspection-to-disposition data flow.
- Defining AI analysis and recommendation boundaries.
- Identifying human technical review points and authority gates.
- Developing synthetic UT inspection scenarios.
- Establishing control points for incomplete, conflicting, anomalous, or materially changed evidence.
- Demonstrating configuration control and controlled workflow changes.
- Creating as-executed records capable of reconstructing what evidence was available, what the AI produced, what authority existed, what human decision was made, and whether execution conformed to the approved baseline.
The goal is to demonstrate AI as an engineered decision-support capability operating inside a controlled technical-authority structure — not as an independent engineering authority.