It knows the door. It never claims the key.
Billions in earned benefits go unclaimed every year across EITC, SNAP, WIC, Medicaid, LIHEAP and Lifeline, largely because the rules are scattered across a dozen agencies. Give Threshold a coarse picture of a situation and it shows you which programs' published rules are worth checking: the rule the office applies, where to apply, and who decides.
Request a Demo See the LineWhy it never says "you qualify"
A screener that prints "you qualify" has quietly made an eligibility decision, and only the administering office is allowed to make one. Get it wrong and a person walks into an agency carrying a false promise. So Threshold is a menu of doors, not a calculator.
| It does | It never does |
|---|---|
| Surfaces programs whose rules are worth checking, by coarse relevance | Says "you qualify," "you're eligible," or "you're approved" |
| Shows the published rule the office applies, cited and dated | Computes a pass/fail against a limit |
| Shows your entered figure next to the cited figure, source shown | Pronounces which side of the line you fall on |
| Names the office that decides, and how to apply | Predicts approval or estimates a benefit amount |
| Routes to free application help | Tells you which program you should apply for |
Built to over-show and never decide
Missing Is the Failure
Relevance keys on situation flags you pick about yourself (works, has children, senior, low income band) and never on a number tested against a limit. An extra door costs a glance. A missed door costs a benefit.
No Model Decides
Program facts live only in a cited corpus, each with a citation, a source URL, an effective date and a last-verified stamp. The screener is deterministic. Its guard blocks any verdict, amount, prescription or comparative conclusion, and fails closed to a neutral routing line.
Two Numbers, No Verdict
Where a program carries a figure table and you enter your own figure, the card sets the two numbers side by side with the figure's citation. It adds no comparator words and draws no conclusion. The office makes the qualify-or-deny call, every time.
Who It's For
Benefits navigators, community organizations and legal-aid intake desks that need a defensible answer to "what should this person go ask about?" that never becomes an eligibility opinion.
Not yet ready for a real person
A 65-assertion self-check verifies the working core, including the guard that blocks verdict wording.
Before this screens a real person
- The corpus lists real federal programs with real citations. Dollar limits change every year and vary by state, so every figure stays SAMPLE until it is verified against the cited source and dated.
- A benefits specialist or legal-aid partner reviews the rule statements and the posture before launch. That review is a named gate.
- If a free-text intake model is ever added, it ships disabled behind a no-training gate.
Watch the guard hold the line.
The live demo runs the real guard on sample data, including the wording it blocks. Fifteen minutes, no slides. Try to make it say "you qualify".
Request a Demo