Your Construction AI Isn't Hallucinating. It's "Faithful but Wrong" — and That's the Scarier Problem

This week we sat down with João Dias, Co-Founder and CEO of Constructer.ai, and Dr Shirin Dora, Senior Lecturer in Neuromorphic Computing at Loughborough University, to unpack what trustworthy AI in construction actually demands — and why "the model made it up" is the wrong thing to worry about.

🎧 Short on time? Hit play — the full conversation is right here.

Everyone frets about hallucinations, the model inventing a fact out of thin air. João and Shirin make the case that the real danger in construction is subtler: an output that is faithful but wrong. The model reasons perfectly over the data it was handed, and still gives you the wrong answer because the data itself was patchy, out of date, or stripped of the context a seasoned project manager carries in their head.

That's why Constructer.ai is built around live data rather than historical averages. Their pitch is a loop, detect, quantify, act, that reads the current state of a project instead of extrapolating from what happened on the last one. Pair that with genuine domain expertise (Shirin's point is that you pick the AI technique to fit the construction problem, not the other way round) and you get insight a team can actually stand behind when liability is on the line.

Adoption is still early. The pair are candid that most of the industry is nowhere near using this well yet, and that the near-term job is building the capability, data discipline, judgement about when to trust a model — before anyone starts daydreaming about AGI running the site.

Why it matters: in construction, a confident, well-reasoned, wrong answer is more dangerous than an obvious hallucination — because someone signs off on it. Trustworthy AI here is about live data, domain expertise, and knowing when the machine is wrong.

👉 Listen to the full conversation with João and Shirin here.

Wispr Flow turns your voice into clean, professional text you can send the moment you stop talking. Not rough transcription you have to clean up. Actual polished text — ready for email, Slack, or any app.

Speak the way you think. Go on tangents. Change your mind mid-sentence. Flow strips the filler, fixes the grammar, and gives you text that reads like you spent five minutes writing it.

89% of messages sent with zero edits. Millions of professionals use Flow daily, including teams at OpenAI, Vercel, and Clay. Works on Mac, Windows, and iPhone.

Start flowing free

Enjoyed this? Subscribe to the Project Flux newsletter and get a step ahead on AI and tech — straight to your inbox every Monday.

Project Flux is independently owned, funded, and operated by its founder. Any external references or features are for community benefit only and do not imply endorsement, control, or ownership.