Skip to main content

​
Resident Intelligence Layer

​
AI Dimension C: The Resident’s Advocate

​
What the Resident Intelligence Layer Is

The Resident intelligence layer is the most user-visible AI dimension in Leja, but it is not a separate product surface. It is Dimension C applied inside Resident: it reads witnessed records, supports human decisions, and makes the Resident surface more useful. Every Resident user has access to support that knows their full Trust Graph — all three tracks — and acts as a trust-aware concierge in everything they do on the platform. The intelligence layer does not just surface matches. It advocates. It knows the user’s full Trust Graph. Their budget signal from Track 1. Their area preferences from search history. Their timeline urgency from their current lease end date. The behavioral patterns of every landlord and property they are considering. And it acts on that knowledge in the user’s interest — proactively, not reactively.

​
Advocate Support for Apartment Hunting (Track 3)

The agent knows:
  • Full Trust Graph tier and what it qualifies them for
  • Budget signal derived from rent progression history + search behavior
  • Preferred areas derived from past tenancies and current search
  • Timeline urgency derived from current lease end date
  • Behavioral patterns of every landlord and property being considered (from the Trust Graph lens matrix)
The agent does: 
Proactively surfaces matching properties:
  Before the user searches, the agent has already identified properties
  that match their Trust Graph tier, budget band, and area preferences.

Flags compatibility mismatches:
  "This landlord has increased rent above inflation every year for
   the past 3 years. Your budget trajectory suggests this would be
   a problem in year 2."

Flags property risks:
  "This property has had 3 unresolved maintenance events in 18 months.
   I'd suggest asking about the building management before committing."

Drafts the application:
  "Your Trust Graph makes you a strong candidate for this property.
   Here is a draft application letter using your verified history
   as the credibility foundation — not a template, tailored to you."

Pre-fills document packs:
  From verified identity data already in the Trust Graph.
  Name, NIN verification, residence history — pre-populated.

Alerts instantly:
  When a property the user was close on changes back to AVAILABLE,
  the agent alerts before anyone else knows to look.

Sets expectations before applying:
  "Your Gold Score makes you a strong candidate for this property.
   Your Budget signal is slightly below the typical tenant for
   this landlord — here is how I suggest framing your application."

​
Advocate Support for Service Requests (Track 2 Consumer Side)

The agent knows:
  • Property maintenance history (relevant to this type of job)
  • Budget band for service work (from overall financial picture)
  • Past provider relationships (who they have hired and how it went)
The agent does:
  • Surfaces compatible providers by trust tier (not just proximity)
  • Flags provider patterns: “This provider has a 68% repeat hire rate from properties with the same type of electrical issue you have described. Strong signal.”
  • Drafts job descriptions that give providers enough context to quote accurately — reducing scope disputes before they start

​
Advocate Support for Leja Stay Booking (Track 3)

The agent knows:
  • Travel context (business trip, relocation search, short-term need)
  • Budget band from Track 1 and Track 3 history
  • Preferred areas from past tenancies and search behavior
  • Trust Graph tier (what hosts will accept)
  • Past Leja Stay behavior (what kinds of stays went well)
The agent does:
  • Surfaces compatible hosts and properties (trust tier matched)
  • Flags property risk signals from the Property RIN
  • Handles booking initiation
  • Monitors check-in and check-out milestones
  • Alerts the user of anything unusual (host late to respond, etc.)

​
Intelligence Transparency Rules

The agent must communicate confidence levels, not just conclusions. If the agent is recommending a property based on thin data (a new landlord with only 1 previous tenancy on record), it must say so: “I have limited data on this landlord — only 1 tenancy on record. Proceed with extra care. Here are the questions I’d suggest asking.” If the agent’s assessment is based on strong data, it says so: “This recommendation is based on 18 months of verified data on this landlord across 4 tenancies. High confidence.” The agent never:
  • Makes a final decision (agent recommends, user decides)
  • Overstates its confidence on thin data
  • Excludes properties from the user’s view without disclosure
  • Acts on behalf of the user without explicit instruction

​
Resident Pro AI Access

Free Resident: Basic AI recommendations (search surfacing, basic flags) Resident Pro: Full advocate capability (all the above, proactive advocacy, pre-filled document packs, application drafting, timeline monitoring) See 07_resident_pro.md for full Resident Pro specification.