#077 - Equipping Engineers with AI for AEC's Realities

For professional engineers, building useful AI tools means custom-fitting them to the messy, context-rich reality of daily work.

The discourse around Artificial Intelligence paints a picture of sleek automation, perfectly defined inputs, and quantifiable outputs. It's based on the idea that we can create perfect assembly lines, where processes are streamlined and every variable is accounted for.

The marketing makes everything look so easy, so simple. This vision holds appeal. Yet, for those of us operating in the trenches, the reality is different.

Our work is rarely about producing identical widgets on a production line. It's actually a relentless, multi-disciplinary exercise in problem-solving with incomplete data, dynamic site conditions, unpredictable weather, evolving subsurface information, ambiguous requirements, and tight schedules. And then there's the human component: Frank's on vacation, Bethany's sick, Paul is at a Coldplay concert and having an affair with his HR girlfriend. These are not edge cases; they are the reality of how we work.

This complex interplay of technical, logistical, and human factors is the true "last mile" problem in AEC, far beyond what any LLM can solve out of the box.

The "Last Mile"

When we talk about applying AI to engineering, the crucial question isn't "Can a model do this?" but "How do we quantify 'good enough' when 'good enough' is a moving target determined by vague client requirements, best practices, available resources, and sheer professional judgment?".

Generally, core business objectives are clear: deliver good quality work efficiently, satisfy client requirements, and organically grow a system that benefits all stakeholders.