Engineering, not memos

Engineers became compliance officers one document at a time

Nobody decided it. It happened one document at a time. Ron Klemencic, chair and CEO of the structural engineering firm Magnusson Klemencic Associates, has watched the codes his profession works under expand "from tens of pages of provisions to hundreds of pages". His name for what engineers risk becoming is blunt: compliance officers.

That is the before-state. An engineer trained to size a beam or route a road spends the week demonstrating, clause by clause, that a design complies with a rulebook nobody can hold in their head anymore. The design itself gets the hours that are left over.

What actually changed: the rules, not the engineers

Be precise about the mechanism. Nobody has measured how a design engineer's week shifted across four decades; that time series does not exist. What can be measured is the volume of rules and required documentation the same week has to absorb, and that curve points one way.

The second generation of Eurocodes does not just revise the old parts; it adds whole new codes, for structural glass and for assessing existing structures, plus a dozen supporting technical specifications. Every addition is individually reasonable. The sum is a corpus no single engineer can carry, applied by the same humans with the same working week. Document volume went up. Human capacity stayed flat. The compliance share of the profession grew in the gap between those two lines.

When the machine holds the rulebook, the engineer gets the engineering back

A rulebook of that size is, above all, a memory problem. And memory is now the machine's strongest suit. An AI system can hold the full corpus, every code part, every national annex, every AMA reference, and check a written deliverable against all of it at once, without fatigue and without skipping the cross-references. What it cannot do is the engineering itself: the deeper analysis, and the judgment of whether this soil, this junction, this client's constraints make the compliant answer the right one. That division of labour is the flip. The machine takes the memory and the rule-checking. The engineer gets back the analysis, the design and the judgment.

There is measured time waiting on the other side. FMI and PlanGrid surveyed nearly 600 construction professionals in 2018 and found 35 percent of working time, more than 14 hours per person per week, going to non-optimal activities: 5.5 hours looking for project information, the rest on conflict resolution, mistakes and rework. That is the mechanical share of the week, the part that exists because documents are produced and checked by hand.

Who gets those hours back matters as much as the count. Junior engineers today spend their first years on formatting, cross-referencing and template discipline before anyone hands them a real design question. When the machine carries that layer, a junior's first years happen at the level they trained for: real problems, real trade-offs, reviewed by a senior with time to explain why. And the seniors Klemencic worries about get to do the thing they were hired away from. They design again.

Amplified, not replaced

The people running large engineering organisations describe the same flip from the top.

The lament was never really about reports. It was about the distance between the work engineers trained for and the work the week actually contained. That distance is now an engineering problem with a solution, and the profession is on the right side of it. "I went into engineering to build things" stops being a complaint about the job. It goes back to being the job description.