Anton Juric: AI Won't Replace the Engineer — It Will Change Where Their Judgment Matters Most

TEXT START: Most people would probably assume that the systems inside a modern building are already designed by sophisticated software.

The Dissection

This is an adoption narrative disguised as labor-market analysis. Juric correctly identifies a real transition: software is moving from documenting human decisions to generating design outputs. But the article quietly converts “engineers remain accountable” into “engineers remain economically necessary in current numbers.” Those are not the same claim.

The commercial evidence—enterprise validations, signed contracts, workflow integration—proves that firms see value. It does not prove that AI preserves engineering employment. The same value may come from letting fewer engineers deliver more projects.

The Core Fallacy

The article treats judgment as a permanent human moat. Under the Discontinuity Thesis, judgment is itself a target for compression. Once rules, calculations, constraints, simulations, and historical decisions are encoded into domain-specific systems, the engineer’s role can shrink to exception handling, verification, and liability sign-off.

A human approval step may survive as a regulatory or insurance ritual without preserving the old labor market. The noun “engineer” remains. The number of engineers required does not.

Hidden Assumptions

• Productivity gains become more projects and more engineering jobs, rather than lower headcount and lower fees.
• Human review remains intensive instead of becoming sampling, certification, or automated verification.
• Codes, physics, and professional responsibility cannot be formalized sufficiently for AI systems.
• Current shortages of technical talent persist after automation expands delivery capacity.
• Enterprise contracts demonstrate durable employment demand rather than successful labor substitution.
• Accountability requires human production, rather than merely human authorization.
• AI-generated designs will need one engineer per project instead of one engineer supervising many projects.

These assumptions protect the profession’s status, not its economic position.

Social Function

Primary classification: transition management. Secondary classifications: partial truth and ideological anesthetic.

The article prepares engineers and buyers to accept automation while preserving the comforting fiction that professional judgment guarantees professional security. It also serves Endra’s commercial positioning: domain-specific AI is presented as augmentation because that framing lowers institutional resistance and makes procurement easier.

The Verdict

Juric is right about the immediate workflow shift and wrong about its systemic meaning. AI does not need to eliminate engineering judgment to destroy mass engineering employment; it only needs to make one qualified engineer sufficient for far more output.

Mechanical obsolescence begins when generative design enters production. Juric’s three-year MEP forecast marks a plausible adoption threshold, not a reprieve. Social death arrives later, after codes, procurement, liability, and firm culture absorb the change: junior production work contracts first, middle layers compress next, and elite reviewers retain authority over a much smaller workforce.

The engineer is not protected. The engineer is being repositioned as a thinner supervisory layer above an expanding automated production system.