The AI tool hunting your next job before you bid it

Cascade raised $3.5 million to do something most construction owners do manually, badly, or not at all, find work before it's on the street. Worth paying attention to, even if you're not their customer yet.

What happened

A startup called Cascade, founded in 2025 by Hannia Zia and Joana Ferreira, closed a $3.5M seed round this month. The pitch: their platform continuously scans industry portals, analyzes client networks, and uses AI to surface projects before they're publicly confirmed. The idea is that by the time an RFP is posted, the relationships that win the work are already in motion. Cascade wants to give contractors earlier signal so they're not always chasing.

Around the same time, Procore announced three new AI-powered "Digital Coworker" packages and a coming feature called Procore Skills, which lets customers embed their own processes and standards into Procore's AI layer. Their framing: the software becomes a kind of institutional memory for your company.

And McKinsey put a number on the broader opportunity, roughly $228 billion in annual value available to U.S. construction through AI and automation by 2030. A big, round, unprovable number, but directionally, nobody serious is arguing construction gets to opt out of this.

Why it matters for your shop

Cascade is interesting, not because the technology is magic, but because the problem it's solving is real. Business development in construction, especially at the $3M to $25M range, is mostly a principal doing favors, attending dinners, and hoping the phone rings. There's no real system. When work slows down, that's when everyone suddenly wants a BD process, and it's already too late.

Whether Cascade specifically delivers is unproven. They're early, the market is fragmented, and predicting private-sector projects before confirmation requires data access that's genuinely hard to build. I'd treat them as worth watching, not worth rushing to sign up for.

The Procore Skills angle is more immediately relevant for anyone already running on Procore. The ability to encode your standards, how you want submittals handled, what your change order language looks like, how you close out a job, into the AI layer is exactly how these tools stop being generic and start being useful. The risk is that most companies don't actually have their standards written down anywhere. You'd need to do that work first, and it's the unglamorous part that determines whether the feature does anything for you.

The McKinsey number I'd mostly ignore. Aggregate potential doesn't tell you what to do Monday morning.

One thing to try this week

Before any AI tool can reflect how your company works, you have to be able to articulate how your company works. Pick one recurring process, submittal review, subcontractor onboarding, change order initiation, and write down the actual steps your best person follows. Not the policy. What they actually do. Keep it under one page.

That document is the raw material for almost every useful AI application in your business. Without it, you're feeding a general-purpose tool general-purpose inputs and wondering why the output doesn't fit your shop. Claude is good at helping you turn rough notes into a clean process doc if you want a first draft fast, but the thinking has to come from you.

One page. One process. This week.

Austin

Source: https://constructiondigital.com/news/cascades-ai-finds-projects-for-construction-companies