The Good Tech Companies - Cascade Raises $3.5M From a16z Speedrun to Predict Construction Projects Before the RFP Exists

Starting point is 00:00:00 This audio is presented by Hacker Noon, where anyone can learn anything about any technology. Cascade raises $3.5M from a 16 Z speed run to predict construction projects before the RFP exists by a Sean Pondy. What if the most expensive project your firm ever lost was one it never saw? Somewhere in a county capital plan, an in-figure building already exists on paper. A bond has been filed. A parcel is quietly changed hands. By the time a request for proposals goes public, the shortlist is often already forming, and the firms that waited for the RFP are abetting for scraps. Cascade, a New York AI startup founded by two former Google and Amazon adjacent operators, has raised $3.5 million to industrialize that head start, and

Starting point is 00:00:47 Andresen Horwitz's Speed Run program is leading the round. The Seed Round, announced today, July Light 21st, 2026, includes ATA ventures, blitzscaling ventures, Indico Capital Partners, Shucker VC, G2C ventures and Snowball VC alongside a 16 Z speed run. Cascade says its customers, firms working on JFK, LaGuardia, data centers and nuclear facilities, have already surfaced more than $10 billion in project opportunities through the platform. A $2.1 trillion plateau changes the growth math. The macro backdrop matters more than at first. appears, U.S. construction spending has been running at a seasonally adjusted annual rate between $2.1 trillion and $2.2 trillion for over two years, and the Census Bureau's latest

Starting point is 00:01:35 C-30 release puts full-year 2025 output at $2,164 billion, down 1.4% from 2024-0.8 billion. The market is enormous, but it is not expanding. That distinction is the entire investment these In a growing market, a firm can ride the tide, in a flat one, revenue growth comes from exactly one place, taking share, which an EEC means winning pursuits someone else would have won. Globally the stakes compound. Oxford Economics forecasts construction work donorizing from $9 trillion in 2022 to $13, $9 trillion by 2037, driven by the U.S., China and India. The volume is coming. The question cascade monetizes as whose is it first. There is a second structural fact working in Cascades favor. Construction remains one of the

Starting point is 00:02:28 least digitized major industries on Earth. McKinsey Global Institute's digitization index ranked U.S. Construction second to last among all sectors, and McKinsey's 2024 productivity follow-up found the technologies that did get adopted mostly digitized existing paperwork rather than changing how work is found are one. Business development, the front of the funnel, as arguably life least digitized function inside the least digitized industry. That is a wedge, not a niche. Sizing the wedge, honestly, investors in the round reference a $900 billion global industry figure, which most plausibly maps to AEC professional services revenue rather than construction volume, since U.S. construction put in place alone runs $2.16 trillion annually. The honest way to size

Starting point is 00:03:13 Cascades opportunity is neither number. The addressable spend is what firms pay to find and win work. proposal teams, BD headcount, CRM and market intelligence tools, and the opportunity cost of pursuits that go nowhere. That pool is large and demonstrably inefficient. SMPS Foundation Research put saverage AEC proposal hit rates at 37 to 44 percent depending on discipline, while Uninette's 2025 AEC-inspire report finds firms winning roughly half of pursued bids, with only 40 percent of firms using any formal go, no-go process at all. Meanwhile, While Corus Docks research from 2025 found most AEC firms attribute 50 to 74% of revenue to existing clients, a dependence that makes new logip pipeline the scarcest asset in the industry.

Starting point is 00:04:01 Every percentage point of win rate improvement, or every pursuit avoided because the fit score said no, is direct margin. Software that credibly moves those numbers prices against revenue, not against a seat license. Reading the paper trail before it becomes a bid, Cascade's core product, the Pursuit Hub, does three distinct jobs, and it is worth separating them because they carry different levels of technical risk. The first job is detection. Cascade continuously reads public and semi-public exhaust, bond filings, building permits, county capital plans, property transactions, earnings transcripts, budget announcements and meeting minutes. Each of these is a leading indicator of construction demand that conventional bid boards ignore,

Starting point is 00:04:43 because bid boards begin at the RFP, the precise moment when informational advantage ends. This layer is defensible mostly through breadth and data plumbing, less through model novelty. The second job is prediction and fit scoring. The platform matches detected signals to firms, estimating not just that a project is forming but which for me's position to win it. This is where the founders stated hedge fund analog yearns its keep. The system treats construction demand the way quant funds treat price formation as something legible in dispersed data before it becomes consensus. It is also where the claims are hardest to verify from the outside, which is why the $10 billion surfaced opportunity figure should be read as pipeline identified, not revenue one.

Starting point is 00:05:26 The third job is the relationship graph. Cascade mines connections already sitting inside a firm's outlook and adjacent software to surface warm paths and tow a forming project. This is quietly the most commercially important layer, because AEC is a referral economy. With half to three quarters of revenue coming from existing relationships, a tool that converts cold signals into warm introductions is selling the industry back its own social capital, structured. The claimed flywheel is that every pursuit run through the platform teaches the system which signals convert and which firms are credible for which work, and that firms increasingly get matched to each other as teaming partners. If real, that is a data network effect layered on a marketplace dynamic, the combination that made

Starting point is 00:06:09 vertical software categories defensible elsewhere. If not yet real, ITIS at minimum a coherent roadmap, knowledge graphs, then bond filings. Founder market fit here is unusually literal. Hania Zia, CEO, and Joanna Ferreira, CTO, met at unlikely AI, the neuro-symbolic AI startup founded by William Tunstilpito, one of the inventors of the technology behind Amazon. on Alexa, which raised $20 million from Amadeus Capital Partners, Octopus Ventures and former Google CFO Patrick Pichette to pursue trustworthy AI. Zia served there as VP of product. Ferreira led the AI platform, building a knowledge graph if world information and training LLM agents to traverse it. Reading fragmented public records into a structured, queriable model of

Starting point is 00:06:56 reality is not a skill they are a learning for cascade. It is the skill they arrived with. The personal layer is the differentiator most AI founders entering construction lack. Both founders come from construction families. Ferreira grew up in a Portuguese town built on construction and carpentry, and Zia's father attempted and lost, a construction business of his own. In an industry where the standard failure mode for outside technologists is misreading how relationship-driven the work is, biographical proximity is a real, if unquantifiable, acid. The go-to-market origin supports it. The first customer, Munoz Engineering, came out of

Starting point is 00:07:32 direct CFO conversations in New York, not a product-led funnel, and the early logo list, firms attached to JFK, LaGuardia, luxury hospitality and nuclear-adjacent work, is disproportionate for a company months into selling. What speed run is actually underwriting? A 16 Z speed run began as a games accelerator and has become a horizontal, sub-minus 1% acceptance program investing up to $1 million per company across two annual San Francisco cohorts. Its presence at the top of this round signals a specific belief that vertical AI in construction is now a speed game, where the first platform to accumulate proprietary pursuit outcome data compounds and advantage later entrance cannot buy. The syndicate composition reinforces the read. Ada Ventures underwrites the founder quality

Starting point is 00:08:19 argument, blitzscaling ventures is explicitly a network effects and product-led growth thesis shop, and Indico Capital anchors the European institutional layer, consistent with Ferreras Portuguese routes and a likely transatlantic expansion vector. The timing of the round against the capital cycle is also instructive. Construction tech funding has been surging, with NYMBL ventures counting $3,000, $7 billion across the first three quarters of 2025, more than double the prior year period, but 80% of third quarter dollars went to post-series A companies, and the biggest checks, permit flows $54 million, Infravision's $91 million, field eyes $405 million, clustered around robotics, permitting and autonomy. Revenue-side software at seed is the thin end of the barbell.

Starting point is 00:09:07 Cascade is raising where the crowd is not. What has to go right? Four things, in rough order of difficulty. First, prediction precision has to survive scale. Surfacing $10 billion in opportunities is a volume claim. The durable metric is precision at the top of the ranked list. because BD teams will forgive a missed signal long before they forgive a week wasted on a phantom project. Second, the data moat has to outrun replication. Bond filings and permits are public. If Cascades edge stops at ingestion, incumbent AEC software vendors with distribution can follow. The proprietary layer must be the outcome data, which firm won what, taught back into the model. Third, the network flywheel has to activate without tripping over confidentiality, since firms feeding

Starting point is 00:09:53 pursuit intelligence into a shared system air also feeding a platform their competitors use. Fourth, a $3.5 million seed buys a finite runway and a sales-heavy vertical. The round is appropriate for the stage, but the next raise will be priced on net revenue retention and win rate deltas, not on the elegance of the signal graph. What to watch for, here is the test this column will hold cascade to. Within 18 months, by early 2028, the company should be able to show, for a cohort of customers, a measured improvement in pursuit win rate or pursuit efficiency against the industry's roughly 50% baseline, and at least one publicly referensible project won from a signal that predated the RFP by six months or more. If those numbers materialize, Cascade is not a lead gen

Starting point is 00:10:38 tool, it is a pricing power business in the least digitized corner of a $2 trillion market. If they do not, the pre-RFP window will remain what it has always been, visible only in hindsight. vested interest disclosure. Hacker Noon has reviewed the report for quality, but the claims herein belong to the author. Hashtag D-Y-O-R. Thank you for listening to this Hacker Noon story, read by artificial intelligence. Visit hackernoon.com to read, write, learn and publish.

The Good Tech Companies - Cascade Raises $3.5M From a16z Speedrun to Predict Construction Projects Before the RFP Exists

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.