Cascade, a New York startup building an AI pursuit platform for architecture, engineering and construction firms, has raised $3.5 million in seed funding led by Andreessen Horowitz’s Speedrun fund. Ada Ventures, Blitzscaling Ventures, Indico Capital Partners, shuckerVC, G2C Ventures and Snowball VC also participated, according to the company’s announcement.
The founders are Hannia Zia, the CEO, and Joana Ferreira, the CTO — both former Google operators who went on to UnlikelyAI, where Zia was VP of product and Ferreira led the AI platform. TechCrunch reports the company was founded in 2025.
The Gap Cascade Is Aiming At
Construction has spent a decade instrumenting everything downstream of the contract. Estimating, scheduling, field capture, quality, payments — each has a well-funded software category and, increasingly, an AI layer. What sits upstream of all of it, the question of which projects a firm goes after in the first place, is still largely run on relationships, trade press and whoever happened to hear something.
Zia’s framing in the announcement is the sharpest statement of the problem: “Every firm we work with can point to a project they should have won but never even had a chance to see.”
That is a real asymmetry. A firm can have the best cost model in its market and still lose on work it learned about three weeks too late, or never learned about at all. The bid is the visible competition; the invitation to bid is the invisible one.
How the Prediction Actually Works
The word “predicts” is doing a lot of work in the coverage of this round, so it is worth being precise about the mechanism, which is less mystical than it sounds.
Cascade reads public signals — bond filings, permits, capital plans, newly announced grants — and infers from them where work is forming before a tender exists. A municipal bond issue for a school district, a capital plan line item, a rezoning: each is a public document that implies future construction, published long before anyone issues an RFP. The company also scores which opportunities a given firm is well positioned to win, and maps relationship pathways to the people involved.
Ferreira described the output to TechCrunch in operational terms: most likely one of these five developers will win this newly announced grant, so go start talking to them.
This is inference from public record rather than prophecy, and that is the honest version of the pitch. It is also the more defensible one. The signals are public but scattered across thousands of jurisdictions and document formats, which is exactly the kind of work that is miserable for a business development team and well suited to machines.
The Traction, and How to Read It
The announcement says customers have surfaced more than $10 billion in project opportunities, and names Smallwood, S.A. Casey Construction, Munoz Engineering and FORGE Architecture among them. Tim Johannesson, a principal at Smallwood, is quoted saying Cascade “helped us respond to and win more projects, including a recent $6M project, that I was only able to find through Cascade.”
Read those two numbers carefully, because they are different kinds of claim. “More than $10 billion in opportunities surfaced” is a pipeline figure — the value of work Cascade flagged, not work anyone won. It measures the funnel’s mouth. The $6 million won project is a single customer-attributed outcome, and it is the more meaningful of the two, though it is one data point offered by the company.
For a seed-stage company this is a reasonable evidence base. It is not yet evidence that the product changes win rates at scale, which is the only number that will matter in eighteen months.
Where It Sits
Cascade is attacking the front of a funnel that several well-funded companies are working further down. PLAN0 is building cost intelligence from project data; Bidflow automates electrical takeoffs; Guthrie AI took $4 million for glazing bids. All of those assume you already know about the job. Cascade’s wedge is one step earlier, and the adjacency is obvious enough that the categories will probably collide — a pursuit tool that knows which projects you should chase is a short hop from a tool that helps you price them.
The competitive question is less about other startups than about incumbents. Construction already has project-lead data services, and they are not new. Cascade’s claim to be different rests on prediction ahead of announcement rather than faster distribution of announcements — a genuine distinction, but one that will need to survive contact with customers who have been sold “better leads” before.
What to Watch
Two things. First, whether the prediction holds up outside the jurisdictions where public data is rich. Bond filings and capital plans are well structured in some US states and close to unusable in others, and a platform that works beautifully in New York and poorly in most of the country has a smaller market than the pitch implies.
Second, whether win rate moves. Surfacing opportunity is the easy half; construction firms are not generally short of things they could bid on, they are short of time and the judgment to pick. If Cascade’s scoring genuinely tells a firm which of forty possible pursuits deserve its three proposal writers, that is a product worth paying for. If it just makes the list longer, it will end up as another feed nobody reads.
