Charles Ma helped deploy crane sensors at Versatile, a startup that raised more than $100m to mount cameras on tower cranes. The lesson he took away had nothing to do with hardware. It was that most construction tools look backward, while the money leaks forward, at buyout, in the scope nobody caught. His new company, Scoreboard AI, goes after that gap, and he’s building it in the opposite style to the hype cycle he watched up close.
Ask a superintendent what they make of most construction software and you’ll hear a version of the same complaint. One of them handed Ma the line that still shapes how he builds: the tool is a rear-view mirror, and I’m trying to look ahead. That stuck. Ma had spent the early days of ConTech inside Versatile, one of the most talked-about hardware startups of its generation, and even there the honest question was whether the product paid for itself.
Ma is now co-founder and CEO of Scoreboard AI, which reads your project documents and works out who is carrying what scope before buyout, the moment a missed note can quietly turn into lost profit. He came on Bricks & Bytes to talk through what a $100m-plus hardware bet taught him, why he has skipped the flashy fundraise on purpose, and where AI actually helps in preconstruction versus where it still falls flat.
Whether you run projects, sell into contractors, or write checks in the sector, there’s a takeaway here for you. Here’s the short version.
Scope gaps at buyout are one of the least glamorous and most expensive problems in construction. Catch them early and you protect your margin. Miss them and you can end up funding the fix yourself. The tools that earn trust are the ones that show their work, and the founders worth watching are the ones ignoring the noise.
The rear-view mirror problem
What a $100m crane-sensor bet actually taught him
Ma joined Versatile in 2017, back when the whole category was young. The deal was blunt: keep your day job, take $16 an hour, and from 6am to 8am be the eyes and ears on site. The first target was the Chase Center, the Warriors arena going up in San Francisco. Versatile went on to raise a seed round, a $20m Series A and an $80m Series B, comfortably north of $100m, to put cameras and sensors under crane hooks and turn a job site into data.
What impressed Ma was the focus. Versatile picked the superintendent as the one stakeholder whose day it wanted to improve, and served that person without apology. Where it ran into a wall was ROI. The output started life as spreadsheets nobody knew what to do with, and proving the tool paid for itself stayed hard. Ma pushed the team toward a timeline view showing who used the crane hook and when, which became a headline feature. But the deeper lesson came from that superintendent’s complaint, and it sent him looking for a different problem to solve.
This tool is a rear-view mirror, and I’m trying to look ahead.A site superintendent, recalled by Charles Ma
Where the margin actually leaks
How a missed note becomes a margin hole
Look at why it’s hard to see ahead and you land on the seam between design and construction. The two are run separately, and the only thing connecting them is a pile of project documents that are hard to read and often wrong or incomplete. A scope gap can be a coordination problem, a missing design, or something simply left off. As Ma puts it, the money is in the notes, and you can’t fix a bad buy. The trouble is these gaps rarely surface until the crew is on site waiting for a fix.
Owen knows the pain firsthand from his estimating days: a note buried on a drawing calls for cast iron, the estimate prices plastic, and by the time anyone spots it the number is baked in. That’s the mechanic behind the 1-10-100 rule, where a change that costs $1 to fix in design costs $10 in construction and $100 or more after handover, a dynamic we dug into in a recent funding roundup. It’s also why pre-con, not the build, is where most damage gets locked in. By Ma’s own rough on-air estimate, which he was quick to flag as back-of-the-envelope, scope gaps can eat anywhere from half to all of a project’s profit margin. On a business that often runs on 2 to 5 percent, that’s the difference between a good year and a bad one.
- It starts small. A note says cast iron, the estimate prices plastic.
- It hides. The gap doesn’t show until the crew is on site, waiting.
- It compounds. Missing scope comes out of contingency first, then profit, then your own pocket.
Show your work
AI that fills the gap, not fakes it
A contractor’s whole job is managing risk, which is why Ma is strict about one thing: you cannot pass along an answer a piece of software gave you, because you’re still on the hook for it. Hand a project team a 300-page report that’s cumbersome or wrong and they’ll lose faith fast, since checking it takes as long as doing the work by hand. So Scoreboard bakes in trust but verify. You upload plans, specs, narratives and geotech reports, the system assigns scope to the right trade, and every claim can be checked at the source in a click. The same logic runs through our look at how AI is automating construction document review: the model is the first pass, not the signature.
Ma is also clear-eyed about what large language models can’t do. Construction runs on rules, on right and wrong, and general models are weak at rule-bound tasks. He points to AI skeptic Gary Marcus and the old example of a model that plays chess but keeps making illegal moves because it has no real grip on the rules. The useful build, in his view, is proprietary logic that fills the gap where a general model stops, a last-mile problem in a different space. It’s why he thinks thin AI wrappers are losing ground: a general contractor with a few sharp people can vibe-code a comparable off-the-shelf tool in-house. What they can’t easily build is the part that captures a firm’s own hard-won lessons and applies them across every future project, not just the one where someone got burned.
Get the construction tech signal, minus the hype
Deals, product teardowns, and founder takes worth your time, every Tuesday.
Join 3000+ ReadersThe anti-hype playbook
Why staying quiet became the strategy
Scoreboard has skipped the splashy raise, and that’s a choice, not an accident. Ma’s read is that a big round comes with a hockey-stick model you’re then forced to hit, which robs you of the freedom to tell a customer the honest truth about what the tool can’t yet do. He points to Quibi as the cautionary tale: enormous funding, a huge launch, and almost nobody there. He even turned down an accelerator’s three-month cadence because a fixed demo-day clock would have pushed the company onto a path at odds with what its customers needed. His pricing philosophy is closer to Costco: run lean, keep margins thin, charge a fair price, and grow through volume, on the bet that software keeps getting cheaper to build.
Discipline seems to be working. After a stint at Founder University, the team retooled around two ideas, go where the pain is worst and put nothing between the customer and adoption, and has since more than tripled its customer base and revenue while signing someone new most weeks. Ask Ma whether AI quietly runs buyout in five years and he says no, then explains why with a window detail: which trade picks up the flashing versus the glazing depends on where the sill terminates, the kind of tacit call you only learn by getting stung. He’s relieved AI can’t do that part. The point of the tool, as he sees it, is to clear the tedious work so experienced people spend their time on judgment. That’s the WIIFM across all three camps: operators get their hours back, founders get a template for building without the noise, and buyers get a tool they can actually check.
| Delivery model | Who’s in early | How scope gets set | Where gaps creep in |
|---|---|---|---|
| Hard bid | Nobody you chose | Bid packages fired out, you own all of a division | You inherit trades you don’t know or trust, and understanding gets outsourced to them |
| GMP / negotiated | PMs, PX, precon, estimators | Detailed scope sheets, bid forms, sub sign-offs, then a leveling exercise | Bid-day war room, subs use different formats, someone forgets to call everyone back |
| Design-build | Major trades (MEPs) | Key trades help shape the design, then the drawing set drops | Drawings can miss what was agreed and add what wasn’t, and finish trades arrive late |
It’s any piece of work that falls through the cracks between design and construction, a coordination clash, a missing detail, or scope simply left off a bid package. Because project documents are hard to read and often incomplete, these gaps tend to stay hidden until a crew is on site waiting for a fix, by which point they cost real money. (Source)
By Ma’s own rough estimate on the show, which he flagged as back-of-the-envelope rather than hard data, a scope gap can eat anywhere from half to all of a project’s profit margin. Given construction often runs on 2 to 5 percent margins, even a modest miss can wipe out the upside, and Owen noted he has seen gaps push jobs into an outright loss. (Source)
You upload project documents, plans, specs, narratives and geotech reports, and the system assigns each item to the correct trade so you can see who is carrying what. It then compiles scope summaries, acceptable-manufacturer lists, plan-versus-spec discrepancies and change narratives, all built to be verified at the source in a click. (Source)
Ma sees it as intentional discipline. A large raise brings a growth model you’re forced to hit, which can push a company to overpromise to customers rather than be honest about trade-offs. He would rather stay lean, keep pricing fair, and grow through delivery and word of mouth. (Source)
Not in the near term, in Ma’s view. Buyout leans on tacit knowledge, like knowing which trade owns a detail based on where a wall or sill terminates, that isn’t written down anywhere. His aim is to clear the tedious document work so experienced people have more time for those judgment calls, not to replace them. (Source)
Because the price of building software keeps falling, a general contractor with a few capable people can now vibe-code a comparable off-the-shelf tool in-house. The durable value sits in proprietary logic that fills the gap where general models fall short on rule-bound work, plus the ability to capture and reapply a firm’s own lessons across projects. (Source)
Related Articles
Pre-Con Is the Bottleneck Nobody Talks About
How AI Is Automating Construction Document Review
Why ConTech Startups Are Moving Upstream Into Pre-Con
