SKUMU IS LIVE — JULY 2026
More than a trillion dollars a year moves through American public programs — housing, energy, infrastructure, grants and incentives — governed by written rules, published scores and records that anyone may inspect but no one can actually read. The logic of those decisions is scattered across regulations, staff reports, scoring sheets, prior rounds and the memory of people who have spent whole careers inside the system.
Tracking the project was a morass of competing platforms and disparate workflows. The important parts of winning the deal were hidden. The attorney assumed the consultant had it. The consultant assumed the public-affairs firm had it. On and on. At Skumu, we fix that.
Skumu reads the record and reconstructs the decision while the deal is still alive.
01 / WHY NOW
The record was always public. It was never readable. The rules that govern a single competitive housing program can run to thousands of pages, and they change every cycle, and behind them sit years of applications, staff reports and award lists that explain how the program actually behaves — a body of evidence no team of humans could hold in their heads, which is why the people who could hold the most of it became the most expensive people in the industry.
AI changed what “readable” means. For the first time, the entire record of how public money moves can be read, structured and held open for inspection — not just summarized, but rebuilt into a working model of the decision itself, reverse engineered to make sure submitted bids are structured to win.
The material was waiting. We structured that record first and got something nobody can shortcut: a head start that compounds with every round the government publishes on deals across industries.
02 / THE RECEIPT
A real $47.6 million affordable-housing application in Los Angeles. We gave Skumu the application and the rules. Not the result.
applicant self-score reproduced by Skumu
the state’s published figure
The two numbers measure different points in the record. Skumu reproduced the applicant's self-score under the written rules. The agency later published its final figure after review. The 4.650-point difference is visible, sourced and explained, which is more useful than a single number with no account of how it was reached.
PROOF BEFORE ANSWER · REPRODUCTION TEST
When we rebuilt California's tax-credit scoring system from the source regulations, we tested it by rerunning published applications from their raw inputs and comparing our number with the number in the public record. The engine reproduces those scores to the decimal, a miss of exactly 0.000. When the record does not support the agency’s math, Skumu does not guess. It declines to answer until the evidence is there.
No proof = no claim. AI suggests. Humans decide.
04 / THE GATE
A project enters the system as a question with real consequences: should this team spend another dollar?
Sherman Way · $47.6M · 31 requirements · 9 evidence items
The operator sees readiness and the cost of continuing.
Self-score final 73.468%
Applicant §10325(c)(9) · reproduced by Skumu
Agency-review delta −4.650 points
CTCAC staff report CA-24-044, p.4
Next move Upload the CTCAC 9% allocation letter
Outstanding in our records · finance
Skumu produces a go, conditional-go or no-go determination, then opens the determination for review. The operator sees readiness and the cost of continuing. Legal sees the governing rule, the source and the unresolved risk. PAP sees the stakeholder, outreach and political work that can still change the outcome. Consultants and grant writers see the competitive position and the work required to improve it.
This is not a dashboard showing everybody the same information. It is one decision record, viewed by the people who hold different responsibilities for making the project real.
05 / THE RECORD
Public capital does not move according to a static rulebook. Programs change, thresholds shift, tie-breakers decide and even the strongest applications can lose to stronger fields. A published award is not just an announcement. It is new evidence about how the decision system behaved.
Skumu studies that evidence through what we call our proprietary Outcome Calibration Loop. Every round an agency publishes makes the record deeper, so the next opportunity is considered against a fuller account of how the money has actually moved. Deterministic rules keep the system anchored to what agencies require. AI makes the growing record usable at a scale no firm has ever attempted. The OCL gets smarter at every turn.
Recursive learning that compounds. The rules are public and anyone can read them now. The structured history of how those rules have actually decided — round after round, program after program — is being built in one place right now.
06 / THE TERRITORY
Affordable housing is Skumu’s first market because it is one of the hardest versions of the problem: dense regulation, competitive scoring, expensive specialists, political judgment and deadlines that do not care how many people were copied on the email. It is also only a fraction of the territory.
Skumu is already running in beta for EV infrastructure, including live projects against California Energy Commission programs. The rules are different, but the work comes down to the same questions: are the costs eligible, is the evidence there and can the project survive review?
A new market means building a new body of rules and outcomes on the same platform. It does not require a new company or an unrelated product. That is how Skumu moves from housing and electrification into education funding, infrastructure, healthcare reimbursement and other regulated markets where public money depends on evidence.
07 / OUT OF STEALTH
Skumu is live at app.skumu.ai, working on real projects with active pilots and new deals entering the platform.
These are real applications reviewed against the governing rules, the project record and the competitive evidence available before an agency makes its final call. Skumu has something more useful than a polished demo: it can show its work.
Skumu sits between a regulated opportunity and the money, time and professional judgment a team will spend pursuing it. The AI builds and maintains the evidence. The decision engine makes the consequences visible. Developers, attorneys, consultants and public-affairs firms use that record to do their own work while the answer can still change the outcome.
If your work moves public capital, bring us a deal.