Launch HN: Fresco (YC F24) – AI Copilot for Construction Superintendents
Superintendents are the busiest and most expensive people on construction sites. Just like doctors in a hospital, supers diagnose and triage issues, make observations about conditions, and delegate action items to their team. Also, like doctors, supers are responsible for an enormous amount of document generation, accounting for the work done in the field so that back office people can keep records for billing purposes.
Right now, supers are taking 30-200 photos per day to document site progress (mostly for liability reasons, or in case of a dispute with the owner over delays and deliverables). At the end of the day, they'll spend a few hours uploading those photos into OneDrive, or project management software like Procore or ACC. They might transcribe those photos; few supers have time to do this thoroughly, so these photos aren’t very searchable. And supers might use those photos to create "daily logs", which are structured documents that account for the work done on site on any given day.
Fresco uses generative AI to compile reports and punch lists for superintendents during their site walks, not after. Supers simply take short videos with voiceover, or upload photos, and Fresco does the rest, creating notes with transcription, photos, assignees, and due dates. Supers can easily text their crew action items, and receive texts back to keep informed of when tasks are done. We also route this to whatever project management software they're using with a single click.
Right now, we charge per site per month, which includes unlimited users on any given construction project. The typical price is $1k/month, but we offer discounts for multi-month or multi-site commitments. If you want to try Fresco yourself, you can sign up for demo access at fresco-ai.com.
We’ve also gotten some interest from other verticals where site walks are common, like commercial real estate. We’re hoping that in the future it may be possible to address these as well.
If you’ve worked in the construction field, we’d love to hear your experiences and insights. We look forward to your comments and feedback!
34 comments
[ 3.5 ms ] story [ 87.6 ms ] threadbedrockwireless.com
Ping me, stephen [at] bedrockwireless.com
* There is a food production company where their QA's do a monthly walkaround. It takes approx. 2-3 hours to type up notes after the walkaround. I'm in the UK and QA's are paid approx. £32k GBP, so 3 hours of their time is more like £50 benefit.
* Lots of logistics companies do daily walks with shift/team leaders. While these aren't usually typed up or anything, it would be great to document them in terms of actions and a tasklist to complete. The alternative to the software would be getting a team leader to write up notes after the walk, and this would take maybe 30 minutes. A team leader might be £28k p.a. so cheaper to get them to do it than buy software at $12k p.a.
The cost of the software would need to be a fraction (e.g. 10%) of what it is at the moment though for these sorts of use-cases to pay off.
Maybe a more generic version of the software not targeted at the construction niche could be something like £49 per month per user? Sounds more like the sort of level I would expect.
But I'm thinking that $1k is like way way way out of the reasonable range of my use case, and this is so different to your current business model I imagine it's irreconcilable.
In my experience, what LLMs, even some of the most advanced ones (o1, Gemini 1.5) are really good at is rationalization after the fact: explaining why they were right, even when presented with direct evidence to the contrary.
I just ran an experiment trying to get various models put footnote references in the OCR of a text, based on the content of the footnotes. I tested 120+ different models via OpenRouter; they all failed, but the "best" ones failed in a very bizarre and I think, dangerous way: they made up some text to better fit the footnote references! And then they lied about it, saying in a "summary" paragraph that no text had been changed, and/or that they had indeed been able to place all references.
So I guess my question is: how do you detect and flag hallucinations?
LLM generated code needs to be read line by line. It is still useful to do that with code because reading is faster than googling then typing.
You can't detect hallucinations in general.
Superintendents usually still check and, if needed, edit/enrich Fresco’s notes. Editing is way faster/easier than generating notes net new, so even in the extreme scenario where a supe needs to edit every single note, they’re still saving ~90% of the time it’d otherwise have taken to generate those notes and compile them into the right format.
https://www.tomshardware.com/tech-industry/artificial-intell...
A better response would have been "we run all responses through a second agent who validates that no content was added that wasn't in the original source". To say that you simply don't believe hallucinations apply to you tells me that you haven't spent enough time with this technology to be selling something to safety-critical industries.
I'm exploring a similar but unrelated use case for generative AI, and in discovery interviews, what I learnt was that site contractors and engineers do not request or expect 100% accuracy, and leave adequate room for doubt. For them, it's the hours and hours of manually writing down a TON of paperwork, which in some industries is often months and months of work written by some of the poorest communicators on the planet. Because these tasks end up consuming so much time, they forgo the correct methodology and some even tend to fill up some reports with random bullshit just so that the project moves forward - in most cases, this writing work is done for liability concerns as mentioned above, rather than for the purposes of someone actually going through it. If the writing part is cleared for many of these guys, most wouldn't have a problem with the reading and correcting part.
I'm sorry if my comment came across as nitpicky; it's just that every time I try to do some actual work with LLMs (that's not pure creativity, where hallucination is a feature) it never follows prompts exactly, and goes fast off the rails. In the context of construction work, that sounded dangerous. But happy to be proved wrong.
Exactly. Oftentimes reports are filled with nonsensical documentation that are only discovered during the discovery process of litigation after a disaster has already happened. For example, from a real safety report at a chemicals facility, there was an instance of a report stating that under high valve pressure "many bad things will happen". Not joking, literally quoted verbatim.
Most companies' legal teams would love to have their engineers write proper documents and most engineers would love to not spend time on documentation. GenAI can fill that gap by at least giving a baseline starting point which can be edited further for a fraction of the time than writing from scratch.
It's not that trades are super complicated in comparison to other fields like web development, it's that there's no GitHub, no source shared among all pros like "here's what I did and how I got it to work." Without a good stack overflow how does the AI judge the quality of workmanship in photos?
You are absolutely right, btw, about google drives and one drives and hundreds of photos and all that. My experience is in dealing with general contractors on smaller jobs, not supers on mega projects, but they have similar issues. Lots of sloppy back and forths and poor tracking of change orders, etc,
What Im trying to say, since I sort of rambled there, is that while processing and sorting and making punchlists is a good idea, I have doubts about AI's current ability to accurately spot code(as in building code, which unlike JavaScript varies by zip code) issues. Does the AI know that you dont have enough clearance at X or does that have to go into the recording?
that's the correct focus, IMO; let the experts be experts rather than pretend that LLMs are all-knowing
nicely done
It is almost always impossible to get someone to repair right away. The supply is nowhere near demand, so it is a problem worth solving IMO.
Anyway, I've been running around compiling and recompiling photos and punchlists, and my reaction was "Coool!"
I'm not your target audience but I have to imagine the people that are would get utility out of this.
If I could make one recommendation: hire a UX/UI designer ASAP. The less technical the audience, the more intuitive and easy to navigate the UI needs to be.
Our company focuses on home service businesses and they get roadblocked super easily. I think you'll be glad you did it earlier rather than later. Otherwise, the ux debt will pile up and it will be quite a project a year down the line.