The archive · AI & Models · Product decision · 2025–2026
Cascade bets AI can predict construction projects before RFPs exist
Cascade scans permits, bonds and budgets to predict AEC projects pre-RFP; a16z Speedrun led a $3.5M seed after it surfaced $10B+ in opportunities.
Cascade
What the business is
Cascade sells architecture, engineering and construction firms an AI platform that scans thousands of early signals — permits, bond filings, budget announcements, meeting minutes — to predict projects before RFPs go out and rank which bids each firm is most likely to win.
Starting capital：$3.5M seed announced 2026-07-21, led by a16z Speedrun with Ada Ventures, Snowball VC, Blitzscaling Ventures, Indico Capital, shuckerVC and G2C Ventures participating (TechCrunch; tech.eu; Tech Funding News).
How it started
Co-founders Hannia Zia and Joana Ferreira met building AI systems at UnlikelyAI, the startup founded by Amazon Alexa co-creator William Tunstall-Pedoe. Both had construction in their families — Zia's father ran a struggling construction business in Pakistan, Ferreira's mother sold materials to builders and her uncle built mansions in the Middle East. Construction executives told them the same thing: firms do not lose work for lack of capability, they lose it because they never learn opportunities exist. They launched Cascade in 2025 and joined a16z's Speedrun accelerator that September.
What happened
Cascade's Pursuit Hub scans bond filings, permits, capital improvement plans, property transactions, earnings calls and local-government minutes to predict emerging projects, then ranks them against each firm's experience and historical win rates. By July 2026 it claimed over $10B in surfaced opportunities and customers including firms behind JFK and LaGuardia airports, Four Seasons hotels and data centers. The $3.5M seed led by a16z Speedrun, announced 2026-07-21, funds go-to-market, industry events and engineering hires.
How it ended up
Still early-stage and scaling: the round funds market entry and hiring, and the competitive field is legacy bid databases such as GovWin IQ and ConstructConnect, which Cascade argues are less AI-native than its prediction engine.
Background
Cascade is an AI startup for architecture, engineering and construction firms, founded in 2025 by Hannia Zia and Joana Ferreira, who met while building AI systems at UnlikelyAI, the startup created by Amazon Alexa co-creator William Tunstall-Pedoe. Both founders grew up around construction: Zia's father ran a struggling construction business in Pakistan, and Ferreira's family sold materials to builders. Their founding observation, repeated by executives they interviewed, was that firms lose work not because they cannot win it but because they never find out the opportunity existed.
Cascade's Pursuit Hub scans thousands of early public signals — bond filings, permits, capital plans, property transactions, earnings calls, budget announcements, meeting minutes — to predict projects before formal RFPs are published, then ranks each opportunity against a firm's capabilities and track record. By July 2026 the company claimed it had surfaced more than $10B in project opportunities for clients including firms that built JFK and LaGuardia airports, with a Smallwood principal crediting Cascade for a $6M project he only located through the platform.
The $3.5M seed, led by a16z Speedrun with Ada Ventures, Snowball VC, Blitzscaling Ventures, Indico Capital, shuckerVC and G2C Ventures, was announced on 2026-07-21 and covered by TechCrunch, tech.eu and Tech Funding News. The founders argue legacy bid databases such as GovWin IQ and ConstructConnect catalog opportunities only after they are public; Cascade's bet is that prediction before the RFP, plus feedback from every win, compounds into an industry map nobody else has.
What has to be true
- The founders had lived the pain: Zia's father could not find enough projects to sustain his construction business in Pakistan.
- Prediction moves the sales start line: firms begin pursuing months before an RFP is public, when bid databases have nothing to sell.
- Ranking opportunities against a firm's capabilities and win history turns a generic lead list into prioritization advice.
- Every won or lost bid feeds back into the model, so the platform gets smarter as customers use it — a compounding data advantage.
- Speedrun's demo-day stamp of approval helped Cascade close contracts with firms behind JFK, LaGuardia and Four Seasons projects.
What can be applied
Sell the moment of maximum leverage: telling a firm which projects exist and which it can actually win, months before rivals see the tender, beats building a better bid database.
Aftermath
As of 2026-09-04 Cascade is using the $3.5M seed to go to market, host industry events and hire engineers. TechCrunch reported customers who built the JFK and LaGuardia airports, Four Seasons hotels and data centers; tech.eu and Tech Funding News put surfaced opportunities above $10B and named the product Pursuit Hub. The company describes itself as based in both London and New York, expanding an AI-native category against legacy databases GovWin IQ and ConstructConnect. The open question is whether prediction quality and network effects hold beyond its first customers.
Sources
- Cascade raises $3.5M to help construction firms find and win projects
- AI engineering project predictor startup Cascade wins backing from a16z accelerator
- Ex-Google, Amazon duo land $3.5M to help construction firms spot deals before they hit the market
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card