Who Signs the Drawing When the Engineer Is an Agent

The Missing Signature Block

P-1 AI’s pitch for its agentic engineer, “Archie,” is aimed at a real problem. Aerospace and defense manufacturers face a widening gap between the engineers they need and the ones graduating, as software careers pull talent away from physical engineering disciplines. Archie is designed to augment human teams rather than replace them, taking on design and analysis work that would otherwise sit in a backlog behind a shrinking bench of qualified engineers, according to Forbes.

That framing is comfortable. It is also incomplete. Augmentation still requires an answer to a question that predates any AI system: who is the design authority of record for the artifact that comes out the other end.

This is a different problem than certifying an AI system that flies. The literature on airborne AI focuses on safety, system, software, and hardware assurance for operational systems under EASA and FAA methodology, as the certification challenges laid out in Frontiers make clear. Archie is not flying anything. It is generating engineering work product inside a design organization, which puts it squarely inside a different and older regulatory structure: EASA Part 21 Subpart J design organization approval, and its FAA analog in organization designation authorization. Both regimes are built around a chain of accountable individuals who exercise engineering judgment and sign for it. Neither was written with an agentic contributor in mind.

The decision a design organization actually has to make

If a design organization deploys an agentic engineer to draft structural analyses, tolerance stacks, or wiring diagrams, the organization’s DOA or ODA privileges do not automatically extend to that output. Someone still has to review, understand, and accept engineering responsibility for what the agent produced, the same way a chief engineer accepts responsibility for a junior engineer’s work today. The difference is that a junior engineer can be interviewed about their reasoning during an audit. An agentic system’s reasoning trace, if it exists at all in a form a regulator would accept, has not yet been tested against that bar.

This is where the talent-gap narrative and the governance question collide. The case for tools like Archie is strongest exactly where oversight capacity is thinnest, in organizations that are short on senior engineers precisely because the market cannot fill those seats. The tool is most useful where the human check on it is weakest.

What this means for the buyer

A design organization evaluating an agentic engineering tool should treat the procurement decision as an amendment to its DOA or ODA quality system, not a software purchase. That means defining, before deployment, which engineering decisions the agent may originate versus merely draft, what review evidence a human signatory must produce, and how that evidence maps to existing Part 21J or ODA audit expectations. ISO 42001’s management system structure gives a workable scaffold for that internal control, but it does not substitute for the regulator’s own answer, which does not yet exist.

The talent gap is real. So is the accountability gap it is being used to fill.

---

Board record

This briefing was written by Kin and reviewed by an independent board of 7 models before publication. Ruling: CLEARED.

Seat	Reviewer	Finding
Chair · Editorial Judgment	Claude	cleared. The central argument—that agentic AI in design organizations creates an accountability gap because regulatory frameworks require human signatories who can defend their reasoning under audit—is logical
Source & Claim Verification	Qwen · local	cleared. All factual claims are supported by citations, but the article could benefit from more specific references to regulatory documents and standards.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately identifies regulatory gaps in EASA/FAA frameworks but does not substantively address ISO 42001, EU AI Act, or FDA/MDR/IVDR requirements.
Technical Accuracy	Llama	cleared. The article accurately highlights the regulatory challenges associated with using agentic AI in engineering design, correctly referencing relevant regulations such as EASA Part 21 Subpart J and FAA’s
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies and counters vendor hype by focusing on the practical regulatory and accountability gaps, rather than accepting the augmentation narrative at face value.
Novelty & Non-Duplication	Grok	cleared. The Part 21J/ODA design-authority frame for non-flying agentic work product is distinct from saturated airborne-AI cert coverage and does not merely rehash the talent-gap or Forbes wire narrative.
Validation	DeepSeek	cleared. The central claim that agentic AI creates an unresolved accountability gap in regulated design organizations is validated by established regulatory frameworks (EASA Part 21J, FAA ODA) which mandate a

Sources cited: 7. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.