Annum lets community organizations, municipalities and employers move a household from need to a leased unit on a shared, consented record — with Dorothy AI proposing eligible options and people making every decision.
Synthetic The demo runs entirely on fictitious households, programs and properties.
Each organization works in its own tenant with its own roles. What flows between them is governed by the member's consent, not by an export.
Household, need, consents, evidence, matches, applications, tasks and communications in one workspace. Dorothy proposes eligible programs and units; your counselors approve, apply and place.
Capacity, spend and the households your funded programs touch — computed only from consented records, with small cells suppressed. Names never leave the delivering organization.
Publish a stipend or deposit-loan benefit with its own eligibility rules, then track enrollment, slots in use, spend and average benefit — enough for the board, nothing about an individual employee.
The same six steps a counselor already takes — with the hand-offs, eligibility maths and reporting done by the platform.
A CBO records the household, its housing need and the consents the member grants to each partner.
Every active rule version and available unit is scored server-side. Criteria are explained in plain language: pass, uncertain or fail.
Proposals sit in a review queue. A counselor approves or rejects; nothing is visible to members or sponsors until then.
An eligible match becomes an application. Unconfirmed criteria become evidence items that clear as documents are verified.
Approval leases the unit, consumes program capacity, retires competing matches and advances the case — all in one audited step.
Funders and employers see de-identified utilization and reach the moment it happens, not at quarter end.
Dorothy scores a household against every active eligibility rule version and available unit a tenant can offer. Each criterion comes back as pass, uncertain or fail with the reason and the sources consulted.
Proposals are held as pending review. They are invisible to the member, to funders and to employers until a counselor approves them — and the approval, the run and the resulting matches share one correlation ID in the audit log.
When a criterion could not be confirmed, it carries straight into the application as a missing-evidence item, so the counselor's next step is always obvious.
Rule set v3 · sources: household, housing need, evidence, unit feed · awaiting counselor review
Every safeguard below is enforced in the platform today and demonstrable in the console.
Every tenant-owned table is policy-guarded by membership. Cross-tenant reads return nothing.
Actor, before/after, reason, correlation ID and source timestamp on every guarded action. The database rejects edits.
HMIS, property feeds and payment rails are reached through backend adapters. The browser never holds a vendor credential.
Dorothy's runs are reviewable, versioned against rule sets, and invisible until a person approves them.
Pilots begin on synthetic data so your team can evaluate the console with nothing at risk, then move to live records under your consent language.