Fight Night 3: Two ConTech CEOs on AI, Architects, VC, and Who Actually Loses by 2030

Fight Night 3 was supposed to be a debate. It mostly was. But somewhere between round one and round three, the two CEOs landed on a thesis that quietly reframes how the entire AEC tech market should be evaluated: owners pay for buildings, not for software, not for design, not for tools. Everything else is a means to that end.

That framing changes the answers. It changes what kind of AI you should buy. It changes who has leverage in the value chain. And it changes whether venture capital is fuel or fool’s gold for the company you’re building.

Below is the unfiltered version of all four rounds, plus the quickfire predictions that lit up the comments section.

Round 1: Should you buy AI from AI companies or industry-specific tools?

The “wrapper” reality and where it actually breaks

Clifton opened with the uncomfortable take. Most of what gets sold as “AI for construction” is a thin layer of prompting over Anthropic, OpenAI, or Google. You’re not buying AI. You’re renting tokens through a middleman who’s paying margin to the same model providers you could access directly. His blunt advice to buyers: open ChatGPT, run a side-by-side test, and see how much of the vendor’s pitch evaporates.

Andrew agreed on the rent-a-model framing but drew the line at workflows. If a tool needs to be adopted across an organisation, with people clicking buttons and integrating it into daily routines, you want a vertical specialist. Not because the AI is better, but because the rollout, training, and risk transfer aren’t a wrapper problem. They’re a services problem. For pure data tasks, horizontal models can be the smarter bet.

Round 2: What happens to architects?

Builders rise, license-holders survive, equity owners get nervous

Andrew’s argument was historical. A hundred and fifty years ago, single entities designed and built buildings. The master builder model held. He thinks we’re going back to it, because owners and developers don’t want designs. They want buildings that solve a business problem at a budget that works. Builders are positioned to take more delivery risk than architects, and the people who carry the most risk usually accrue the most power.

Clifton went the other way. In a world where AI and robotics commoditise labour, the constraint shifts from execution to professional accountability. Architects are the only humans in the stack with a license, and they’re trained to think across silos. That makes them well-suited to orchestrate AI agents pulling information from structural, MEP, civil, and cost domains. Near term, he agrees with Andrew: execution wins, and most architects end up working inside builders or developers. Long term, the licensed human becomes the last responsible party in a heavily automated stack.

The market is already showing the strain. AutoCAD adopters wiped out blueprint firms, but the architecture market still fragmented rather than consolidated. The barrier to opening your own firm is too low for tech alone to drive consolidation. The bigger story is who captures the value: the partner-track billable hour model leaves no room for systems thinkers, who tend to exit into software companies. That talent leak is the real industry problem.

Round 3: What does AEC look like in 5 years?

The owner gets leverage. The middle gets squeezed.

Andrew’s thesis is the cleanest articulation of how to read the next five years. Owners want buildings, not software. They are already used to buying architecture services, project management, subcontractor work. Any technology that makes those services cheaper, faster, or more predictable for the owner gets pulled into the value chain. Any technology that sits outside that and tries to sell its way in faces a long, expensive slog.

Clifton’s framing is compatible but adds a twist. Soft services (architecture, engineering, owner’s rep, permitting) have never had access to nearly infinite labour. AI gives it to them. The brilliant individual practitioner who used to be capped by hiring and training overhead now scales without it. That’s bad news for the standard firm structure. It’s potentially excellent news for the systems thinkers who never wanted to run a partnership-track practice in the first place.

Round 4: Venture capital, fuel or fool’s gold?

Three IPOs in 15 years. Do the maths.

Andrew laid out the structural problem plainly. There have been three AEC IPOs in the last 15 years: Aconex, Procore, and EquipmentShare. True venture investors underwrite billion-dollar-plus outcomes. The base rate in AEC for that scale of exit is brutally low, which means VC dollars chasing the wrong companies create misaligned incentives. Founders raise. The path narrows. Customer needs and investor needs diverge. Companies shut down. Customers get left in the cold.

Clifton’s experience was more personal. TestFit was at $200K ARR working hundred-hour weeks when his co-founder forced the conversation about raising. They picked investors who knew construction over investors with the highest valuation. Even with the right board, the reporting overhead doubles or triples. The board exists to manage your worst instincts. If you don’t need that, you probably don’t need the money.

Both founders said they’d raise again. Both said they’d raise later than they did. The lesson isn’t anti-venture, it’s anti-mismatched-venture. Match the profile of return your company can credibly hit to the profile of return your investor is hunting. For some problems, that’s a venture-scale outcome. For most, it isn’t. Look at IKEA. Look at Esri. Bootstrapped, durable, world-changing.

Quickfire: Who loses by 2030?

And other predictions worth pinning to the wall

The quickfire round produced the line that lit up the comments. Clifton picked owners as the biggest loser by 2030. Andrew nearly broke character: how could owners lose when the entire industry is getting more efficient at delivering the only thing they actually want? They’re the ones turning a business problem into a building. If buildings get cheaper, faster, and more predictable, owners win.

On the next IPO, both were cautious. Andrew floated Buildup as a name with momentum, but flagged the bar is now extremely high. Clifton predicted whoever IPOs next will be a services firm, not a pure SaaS company. An AE firm deciding to be AECOM 2.0 with AI baked in. On the long tail of smaller contractors, both agreed: some will be in trouble, some will digitise, many will keep five owner relationships and quietly run profitable businesses for decades. Diversity, not death.

Comparison: When to buy what

Frequently asked questions

Related Articles
Nemetschek Acquires HCSS: Why the Infrastructure Bet Is the Biggest Signal in ConTech
Lessons From Levelset’s $500M Exit to Procore
Primepoint Raises $10M Seed to Build AI That Reads Construction Drawings