Roadmap — from AOZ pilot to an integration operations system

Related: Changelog · Blog · AOZ pitch · Integration research framework

created_date: 2026-01-15 last_modified_date: 2026-08-19 last_modified_summary: Reframed the roadmap headline and thesis around integration operations, with housing as the base layer.

The thesis

Housing shortage in Zürich, New York, San Francisco — and weakening family structures everywhere — mean more people who are not related to each other sharing territory: rooms, apartments, buildings, subdivisions. Wherever two people share territory, there must be a common understanding of how living together works. Today that understanding lives in gut feeling, paper Hausordnungen, and staff who firefight conflicts after they explode.

The product is an integration operations system for Fachpersonen with housing stability as the base layer: who lives where (placement & vacancy), who fits whom (compatibility matching), what the rules are (layered governance), what happens when stability breaks (conflict ladder, maintenance, incidents), and how language, capability, participation and coordinated guidance progress over time. AOZ asylum housing is the first vertical — deliberately the hardest one: vulnerable residents, high conflict cost, strict privacy duties, and published research to build on. A system that works there works anywhere.

The scientific basis is now broader than cohabitation alone: see docs/INTEGRATION-RESEARCH-FRAMEWORK.md for the product's five research pillars (housing stability, language/orientation, capability/work readiness, participation, and coordinated guidance).

Not the mission: maximizing occupancy or throughput at the cost of wellbeing. Reduced conflict is the business case (see CLAUDE.md ROI figures); dignity, transparency and data minimalism are the moat, not the constraint — property managers buy risk reduction, and residents only feed the matching system honest data if it never punishes honesty.

What generalizes (and already does)

The architecture bet is that verticals are config packs, not forks:

Layer	Mechanism today	Generalization
Branding	BRAND presets (aoz/aozh/wg)	Org-level theming, orgName separate from product
Rules	ORG_RULE_CATALOG + delegation (FIXED / UNIT_MAY_STRENGTHEN / UNIT_DECIDES)	Any org's rulebook is a catalog seed; the delegation hierarchy models campus→building→unit→room
Matching	Factor configs with weights (resident-factors.ts)	Per-vertical factor packs (student, senior, coliving)
Space	Unit → room → spot hierarchy	Shared rooms (student/army), whole units, buildings, subdivisions
Money	Pure Rappen-integer expense engine	Currency is a formatting concern; engine is portable
Reports	Category → desk routing	Desk set is org config

Horizons

H1 — Win the AOZ pilot (now → Q4 2026)

The only thing that matters: prove the metrics (incidents −30%, relocations −50%, mediation hours −40%) at 1–2 locations against the baseline. Everything in H1 serves that.

• ✅ Real deployment (Witikonerstrasse 458), demo doors, governance, expenses, report routing.
• ✅ Signed Hausordnung as the in-app org rule catalog, with per-version acknowledgement — done (see changelog).
• ✅ AOZ vs WG surfaces (nav, expenses/votes, code-first login, fast matching).
• ✅ Staff roles: Leitung, Betreuung, Sozialarbeit, Jobcoach.
• ✅ Learning records (tests, courses, informal) for residents and Jobcoach.
• ✅ Learning achievements and volunteering/community-service hours; language offers reuse Aktivitäten.
• ✅ Role-shaped integration evidence boards: overview, Jobcoach, and Freiwilligenarbeit with filters for own clients, source, status, and category.
• ✅ Care team (Wohnen, Sozialarbeit, Jobcoach) visible to the resident.
• ✅ Care workspace: appointments and catalog attributes per domain; one login can be resident + Leitung.
• ✅ Portal chrome: collapsible sidebar + header account/language; dead Wohnung/Mitbewohner pages removed.
• ✅ Room-level matching and a short intake as the default.
• ✅ Complete locale packs offered (AR, FA, TI, UK, RU, TR) plus DE/EN/FR vouched.
• ✅ Baseline/outcome dashboard on the staff home and analytics.
• Close the loop from incidents to matching: recorded conflict outcomes feeding the risk dimension per factor pair (transparent weights, no black box — every score explainable in German). Wait for AOZ outcome data.
• Acknowledgement coverage as the leading indicator on the staff dashboard.

H2 — Multi-tenancy without forks (2027)

One deployment currently serves one org. The multibillion path requires one platform serving many:

• Organization as a first-class row (not env config): rule catalog, contact sheet, factor weights, brand, locale per org. organization.ts and house-rules.ts become per-org seeds — the code paths already read them as data.
• Property hierarchy: org → property → unit → room → spot, with rule delegation at every level (a building Hausordnung between org floor and house rules — the AOZ paper document already implies this level).
• i18n: resident chrome has locale packs (DE/EN/FR vouched; AR, FA, TI, UK, RU, TR complete). Swiss High German stays the binding Hausordnung text. Incomplete files (sq, so) stay hidden.
• Vertical factor packs: student housing (semester rhythm, study/party axis), coliving (profession, guest frequency), senior co-housing (care-adjacent functional needs — never diagnoses).
• Self-serve onboarding for small operators (a WG can already run it — the live instance proves it; make that a product, not a favor).

H3 — The network (2028+)

• Resident-consented portable profiles: preferences travel with the person across orgs (they, not the org, own their compatibility profile).
• Marketplace/community layer: exchange, shared purchases, skills — the expenses engine plus the community feed make retention.
• Outcome dataset as a public good: anonymized, aggregate conflict/ compatibility research back to the field the product came from.
• Vacancy network across operators in one city — minimizing vacancies city-wide, matching people to homes, not just beds.

Engineering invariants (any horizon)

1. Verticalization by config, never by fork. If a new customer needs a code change beyond a config pack, the architecture lost.
2. Every score explainable in the resident's language. No black boxes — trust is the product.
3. Track functional needs only; never diagnoses, immigration status, religion, politics. Data minimalism scales trust across verticals.
4. Money is integers, balances are computed over full history, votes snapshot their policy — auditability everywhere a dispute can happen.
5. Safety is never put to a vote, and safety conflicts never enter the ladder at the bottom.