Construction’s Great Stagnation
Thoughts from ‘Boom: Bubbles and the End of Stagnation’
Over Christmas I’ve been reading Boom: Bubbles and the End of Stagnation by Bryne Hobart and Tobias Huber.
In it, the Great Stagnation is framed as a slowdown in real progress: fewer step-changes, more marginal gains, more optimisation theatre.
Seen through a construction lens, the diagnosis is brutal.
The Great Stagnation — Construction Edition
Construction hasn’t stagnated because we failed to digitise. It stagnated because we digitised the wrong things.
We swapped paper for PDFs, drawings for models, folders for CDEs - but we are leaving the production system relatively untouched.
What actually happened:
Productivity flatlined for 30+ years {source}
Projects got more complex, not faster
Compliance expanded faster than capability
Risk was pushed downstream, not removed
“Innovation” became software procurement, not system change
BIM didn’t change how we build, it changed how we describe what we intend to build. That’s not real progress; that’s better documentation of inefficiency.
Whilst being able to coordinate design and review buildability more easily has brought some improvement to planning the build, there have been very few examples of progress beyond the preparation and information gathering phase.
The industry optimised coordination costs while ignoring throughput, variance, and decision latency - the real killers.
That’s where constructions stagnation lies.
Explainer: Variance, Throughput, and Decision Latency
Variance = The gap between what was planned and what actually happens — in time, cost, quality, or sequence.
Throughput = The rate at which the system converts inputs into completed, accepted work.
Decision Latency = The elapsed time between a decision being needed and that decision being made and acted upon.
If we can accept this then what would actually count as real innovation?
Three Opportunities for Genuine Progress
Industrialised Construction (Not “MMC”, the real thing)
Industrialisation collapses three constraints at once:
Variability: the delta between plan vs reality is narrowed
Labour dependency: the workforce can be mechanised more readily
Learning rate: lean and continuous improvement can be more easily implemented
A factory learns. A project forgets.
When work moves from site to factory the output becomes repeatable, quality variance collapses, and productivity compounds instead of resetting to zero every job.
This is why manufacturing got rich and construction didn’t.
But we have MMC, right, what went wrong?
We tried to finance MMC before we standardised it
Every “platform” was bespoke in disguise
Clients demanded uniqueness; factories need sameness
To make genuine progress Industrialised Construction needs to look towards:
Brutal standardisation of interfaces
Fewer typologies, not more
Platforms treated as products, not projects
If it doesn’t feel uncomfortable to architects and clients, we’re not industrialising enough.
Decision Automation at the Production Edge
Construction isn’t slow because people are lazy. It’s slow because decisions queue.
Everyday, on every project, people are waiting for information, approval, clarification, or assurance.
Most “digital” tools just move the queue onto a screen. To generate progress here we must address the queue, by:
Pre-authorising decisions
Embedded deterministic rules
Automate compliance checking
Require human intervention only by exception
Some examples of this are:
Automated permit validation
Deterministic design rule checking before issue
Real-time quality acceptance at point of install
Commercial rules enforced upstream, not after the fact
If an excavator is standing still waiting on a decision or information, the system is broken.
Information as an Asset, Not a Deliverable
Construction treats information as something you hand over, not something you compound.
Every project relearns the same lessons, rebuilds the same spreadsheets, rewrites the same risk registers, and loses most learning at close-out.
That is commercial suicide.
The real innovation is treating information like capital:
Structured once
Reused many times
Queried, not searched
Fed back into future bids, designs, and methods
This is where things like:
Knowledge management (perhaps through knowledge graphs);
Capturing and recoding production data; and,
Measuring cost, time, risk, and quality signals
become a competitive advantage, not a compliance burden.
Businesses that do this will price risk better, build faster, and quietly destroy competitors who still “start fresh” every job.
But Wait… Where are the Robots?
Robotics matters - but not in the way most people sell it.
Right now, robotics in construction sits in the same bucket BIM did in 2012: over-promised, under-integrated, and mostly bolted onto a broken production system.
Let’s look closer.
The Role of Robotics in Construction (No Hype)
Robotics is not the next excavator moment. It’s a force multiplier after you fix three preconditions:
Standardised work
Deterministic sequences
Digitally controlled decisions
Without those, robots just expose chaos faster.
Where Robotics Actually Creates Progress
Variance Destruction (The Real Prize)
Robots don’t get tired, distracted, or “interpret” drawings.
They execute the same task, to the same tolerance, every time.
That matters because variance is the hidden tax in construction:
Rework
Snagging
Claims
Programme drift
Robotics attacks variance at the source - in execution predictability.
This is why the first real wins are boring:
Rebar tying
Drilling
Surface finishing
Not Tesla humanoids. Not general labour. Precise repetition.
Labour Substitution Is Secondary
The lazy narrative is “robots replace people”.
The real impact is:
Robots replace scarce skill
Humans shift to supervision, setup, and exception handling
In practice it’s an augmented future. One skilled operator + robot outperforms five non-robot augmented tradespeople.
This is how construction can survive the labour threats we face today - our ageing workforce and skills shortages (compounded by immigration volatility).
Robotics isn’t anti-people. It’s anti-fragility to labour shocks.
But…Robotics Only Works with Industrialised Inputs
Here’s the part the industry avoids: Robots need predictable geometry. This means:
Standard components
Known tolerances
Clean digital models
Stable sequences
Which is why site-based robotics have struggled for traction, whilst Factory-based robotics have scaled.
If your design still allows every detail to be “worked out on site”, robotics will never scale.
What Robotics Is Not Good At (Yet)
Robotics has current limitations, they struggle with:
❌ Highly bespoke work
❌ Constantly changing environments (until we have World Models at scale - post in the works and coming soon!)
❌ Ambiguous scope/task
❌ Poor information fidelity
❌ Adversarial contracts/incentives
Unfortunately, construction loves to manifest all five of these limits!
Until scope, risk, and incentives align, robotics adoption will remain local, not systemic.
The Strategic Role of Robotics
Robotics is not the primary innovation. They are the accelerant.
We can think of it like this:
Industrialisation to create repeatability ➕ Decision automation to remove friction ➕ Information compounding to compound intelligence ➡️ Robotics exploits all three at machine speed.
Introduce robotics too early and you get expensive demos. See various examples of ‘Spot the Dog’.
Introduce it at the right moment and you get step-change productivity.
The Hard Truth
Robotics will not save construction firms that:
Refuse standardisation
Worship bespoke design and execution
Run projects as one-off adventures
Treat data as paperwork
But robotics will concentrate competitive advantage disproportionately to the few. A small number of firms who are able to fundamentally transform to beat variance, decision latency, and intelligence scaling will:
Build faster with fewer people
Control risk upstream
Absorb labour shocks
Win on margin, not heroics
Everyone else will call robotics “not mature yet”.
The Uncomfortable Conclusion
Construction’s stagnation isn’t technological. It’s organisational and incentive-driven.
Fragmentation, Path Dependence, and Misaligned Incentives are the three horsemen of constructions stagnation.
Real innovation reduces variance, collapses decision time, and increases throughout.
Anything else is productivity simulation.
A simple question to ask as a litmus test is:
Does this innovation let us build the next project faster, cheaper, or with fewer people — without heroics?
If not, it’s not progress. It’s noise.