Three conversations. One uncomfortable through-line. Martin Fischer says the industry is investing in AI back-to-front. Suffolk's CTO proves it. Two AEC software founders disagree sharply on who inherits the industry when it finally gets the sequencing right.

THE EXECUTIVE BRIEFING
THIS WEEK’S KEY TAKEAWAYS

Full episode write-up at the bottom↓

Most firms are spending the wrong 80% on AI. The data sitting in your systems was produced by workflows that won't exist in five years. Train AI on that exhaust and you're automating the past, not building toward it.

Even the most data-mature contractor in the United States is an edge case. Suffolk Construction's CTO said it plainly: the gap between the frontier and the field is wider than most executives realise.

The consolidation-versus-fragmentation question just got sharper. Two AEC software founders looked at the same evidence this week and reached opposite conclusions about who inherits the industry. Both have a case. You need your own perspective.

7 THINGS WORTH YOUR ATTENTION
ON THE RADAR THIS WEEK

POWERED BY:

A Stanford professor of 40 years walked into a room of London's most senior construction and technology executives on Tuesday night and opened with something designed to be uncomfortable.

If you have a hundred dollars to spend on AI in your business right now, spend twenty cleaning up the data you already have. Spend eighty building completely new digital workflows that produce data as a by-product.

Twenty and eighty. Most firms in this industry are doing it the other way around.

Professor Martin Fischer has spent four decades at Stanford building the foundational thinking behind virtual design and construction. His argument cuts past the hype question entirely. The industry is sequencing its investment in the wrong order, and the firms that get this wrong won't be able to catch up once competitors compound their advantage.

His analogy is blunt. It's a bit like FedEx showing up while you're still optimising the performance of your Wells Fargo stagecoaches. You can get better stagecoach data. It will not help you compete with FedEx.

Fischer's most actionable point is the one least likely to appear in a vendor pitch. Every person in your organisation should own one digital feedback loop by the end of this year. One loop, owned by an individual, with a name attached to it.

The examples are unglamorous. You said the energy performance of this building would be X. What was it actually? You said this structural detail was constructable. Was it? Did anyone go back and check?

The industry moves forward. Design, build, hand over, move on. Almost nobody closes the loop. Without those loops, you can't improve the workflow. Without improving the workflow, you can't generate useful data. Without useful data, AI is an expensive way to automate the past.

Build the loops in 2026. Layer the AI on top in 2027.

That's Fischer's ordering, and it's the opposite of how most contractors are sequencing this right now. The warning that sticks is this: if your competitors build the new workflows and you don't, you cannot catch up. Their learning curve compounds and yours doesn't.

Jit Kee Chin, Chief Technology Officer at Suffolk Construction in Boston and co-founder of Suffolk Technologies, has lived this. Suffolk hired a Chief Data Officer years before most contractors had heard the term. They built the data infrastructure, the dashboards, and drove operational performance changes. They've done the work most of the industry is still describing in strategy documents.

And here's what Dimitri Stefanescu, founder of Speckle (which Suffolk Technologies has just invested in), said when asked how Suffolk compares to a typical customer: Suffolk are an edge case they're learning from. They represent the frontier.

The most data-mature general contractor in the United States, by general consensus, is a firm nobody else can replicate today.

Even Suffolk, having done all of that work, is now saying the next move isn't another point solution. It's infrastructure. Specifically, the ability to take design data out of the file format prison it currently sits in and make it usable at the object level so that real automation becomes possible.

Jit Kee said it directly: the era of investing in tools that just digitise an existing workflow is largely over. What you still need is the infrastructure to get your data in good order and get it back out.

Which is exactly Fischer's point, said by the buyer instead of the academic.

Andreessen Horowitz (a16z) manages over ninety billion dollars. Their portfolio includes Facebook, Airbnb, GitHub, and Coinbase. Earlier this year they raised fifteen billion in fresh capital. And this week they published a thesis saying construction is the next big bet.

They've made bets in this space before. What's different this time is the public flag-planting. When a firm like a16z publishes a thesis like this, they're doing more than sharing their thinking. They're putting up a bat signal to founders, saying openly: build here; we want to write cheques.

Fischer, asked at Tuesday's event how London compares to Silicon Valley, told a story about a Norwegian friend who walked into an Apple Store in Palo Alto to fix his phone and two hours later had co-formulated a credible startup with the woman helping him.

That density, that speed, that willingness to build something with a stranger in two hours: that's the cultural muscle Silicon Valley still has. And a16z deciding that construction is worth their attention is that muscle being applied to this industry in public at scale.

The practical call to action from this week's a16z episode came from Joe: pay for the advanced LLM subscription. Use it every day for two weeks. Develop a personal sense of what's possible. Then figure out where it fits in your business. His warning was simple. The firm that figures out how to use it will be more competitive than yours.

The third conversation this week produced the sharpest disagreement the Bricks & Bytes Fight Night format has run.

Clifton Harness, founder of TestFit, and Andrew Zukoski, founder of Join, are two venture-backed AEC software founders who looked at the same evidence and reached opposite conclusions about where this ends up.

Here's how the core divide breaks down:

Andrew's thesis (consolidation): Owners want buildings. Builders are best positioned to take risk and deliver end-to-end. The gravitational pull will be toward design-build entities that can execute the whole thing. The master-builder model returns. Architects, over time, work for builders.

Clifton's thesis (fragmentation): AI gives the brilliant individual access to infinite labour. The standard-deviation-ahead operators will compound. The architecture profession fragments, but in a good way, with licensed humans becoming more valuable as the last responsible signatures in an automated stack.

Both are happening simultaneously in the data. Suffolk vertically integrating into design is Andrew's thesis playing out in real time. a16z funding the infrastructure for the brilliant individual is Clifton's thesis playing out in capital.

The most practically useful line from the whole hour came from Clifton on evaluating AI vendors: A lot of the AI tools being sold to construction companies right now, you could open ChatGPT and get 80% of the value in 1% of the time, with zero of the headache. Pair that with Andrew's framing, which was that workflow tools should be vertical-specific while data problems can be horizontal. That's a usable mental model for any executive sitting through a vendor pitch this quarter.

One more thing from Andrew worth holding. There have been three IPOs in the AEC space in the last 15 years: Aconex, Procore, and Equipment Share. Venture capital is hunting for billion-dollar outcomes, and this industry produces them very rarely. His warning to founders raising money was direct: you don't put rocket fuel in a jet. If your business isn't built for the return profile a venture is hunting for, taking venture money will accelerate you into a wall.

It's worth holding that next-to-last segment's a16z optimism. Both are true at the same time.

One. Pick one digital feedback loop in your business. Assign it to a named individual. Set a date. That's Fischer's homework, and it's the highest-leverage item on this list.

Two. If you don't have one, pay for the top-tier subscription to ChatGPT, Claude, or Gemini this week. Use it every day for two weeks. You can't lead a digital strategy from the sidelines.

Three. Next time a ConTech vendor pitches AI tooling, ask one question first. What percentage of this could I get by opening ChatGPT and writing a good prompt? If they don't have a sharp answer, that tells you what you need to know.