AI in Engineering: Finding the Right Balance

AI is neither magic nor useless in engineering. The real value lies in combining deterministic engineering systems with AI orchestration — and knowing exactly where the line between them sits.

The problem with the extremes

Two positions dominate the current conversation, and both are wrong.

Over-reliance: assuming models can solve complex engineering challenges through a well-worded prompt, skipping the structure and validation that engineering exists to provide. This is how you get confident nonsense signed into a design.

Complete rejection: dismissing AI entirely in favour of only hard-coded, deterministic systems. This preserves rigour and forfeits leverage — and the leverage is real.

The framework that works

The practical model sits in between, with a clear division of labour:

• Critical calculations remain hard-coded and deterministically validated
• Those calculations are exposed as tools with strict input validation
• AI orchestrates workflows — it never replaces core engineering logic
• Guardrails stop invalid inputs before anything executes
• Human review keeps accountability where it belongs

The model decides when to run the calculation and what to do with the result. It never decides what the physics should be.

Where the discipline goes

The hard part isn't the AI — it's the engineering judgement about where determinism is essential versus where orchestration creates value.

Calculation of a lifting arrangement: deterministic, validated, non-negotiable. Assembling the inputs, checking the document trail, drafting the report around the result, chasing the missing data: orchestration. That split has to be made deliberately, system by system.

Closing thought

AI will play a role in the future of engineering — but not as a replacement for engineering itself.

Getting this right requires discipline, structure, and strong engineering practice. There is no shortcut, and the teams looking for one are the ones who will get burned.