UnitCycle AI — Development Plan

Overall Progress

Phase 1 — Foundation

0 / 0

Phase 2 — Revenue

0 / 0

Phase 3 — Intelligence

0 / 0

Phase 4 — Delight

0 / 0

Task Tags

Backend API / Django

Frontend Angular UI

AI/ML Model / Prompt

Data DB / Migration

Design UX / Mockup

DevOps Infra / Deploy

Test E2E / QA

Blocker Prerequisite

S 1-2 days

M 3-5 days

L 1-2 weeks

XL 2-4 weeks

▼

Phase 1 — Foundation: Lifecycle + Renewals + Collections

▼

⚙

1.1 Lifecycle State Engine

L 0/0

Data & Models

Add lease_status field to TenantLease model: active, month_to_month, notice, expired, pastData

Create Django migration for new lease status field + update unit status choicesData

Write management command to reclassify 1,146 expired leases (60% MTM, 25% notice, 15% vacated)Backend

Add holdover_start_date, notice_date, move_out_date fields to TenantLeaseData

Create daily Celery task to auto-transition leases: active → expired when lease_end_date passesBackend

API

Update lease serializer to include new status, holdover days, days-since-expiredBackend

Add POST /leases/{id}/convert-mtm/ endpointBackend

Add POST /leases/{id}/give-notice/ endpoint with move_out_date paramBackend

Add POST /leases/{id}/mark-vacated/ endpoint (updates unit to vacant)Backend

Add GET /properties/{id}/lifecycle-summary/ — counts per stateBackend

Frontend — Unit Detail

Design 4 state badges: Active (green), MTM (blue), Notice (orange), Vacant (gray)Design

Replace single "Occupied" badge with lifecycle-aware badge in unit headerFrontend

Adapt lease progress bar per state: normal / "Rolling" / countdown / hiddenFrontend

Show state-specific action buttons (Convert MTM, Offer Renewal, Issue Non-Renew, Begin Move-Out)Frontend

Add "AI Recommendation" card below header for expired/MTM statesFrontend

Frontend — Property View

Update honeycomb hex colors: green=active, blue=MTM, orange=notice, gray=vacantFrontend

Update tooltip to show lifecycle state instead of just "Occupied/Vacant"Frontend

Add lifecycle summary bar to property header (pie or stacked bar with counts)Frontend

Testing

Playwright E2E: verify 4 badge types render correctly on unit detailTest

Playwright E2E: verify lifecycle actions (convert MTM, give notice) work end-to-endTest

▼

🤖

1.2 Renewal Intelligence Agent

XL 0/0

Depends on: 1.1 Lifecycle State Engine

AI / ML Core

Design churn prediction feature set: payment history, maintenance count, tenure, rent gap, sentimentAI/ML

Build feature extraction pipeline from TenantLedgerEntry + UnitsMaintenance + TenantLeaseAI/ML

Train churn prediction model (XGBoost or similar) — or use LLM-based scoring with structured promptsAI/ML

Build rent optimization algorithm: market rate, tenant sensitivity, vacancy cost, comparable unitsAI/ML

Build renewal timing optimizer: when to send offer for max acceptance probabilityAI/ML

Create renewal letter template with AI personalization (tenant name, tenure, improvements, incentives)AI/ML

Backend

Create RenewalRecommendation model (lease FK, churn_score, recommended_rent, optimal_send_date, status)Data

Create Celery task: nightly scan of leases expiring in 30-120 days, generate recommendationsBackend

Add GET /renewals/recommendations/ — list with filters (property, urgency, status)Backend

Add POST /renewals/{id}/send-offer/ — marks as sent, logs activityBackend

Add POST /renewals/{id}/generate-letter/ — returns AI-generated renewal letterBackend

Frontend

Design renewal recommendations list page (table with churn score, rent suggestion, actions)Design

Build renewal dashboard component: pending offers, sent offers, accepted, rejectedFrontend

Add churn risk badge to unit detail header (green/amber/red with percentage)Frontend

Build renewal offer preview modal (shows letter, rent options, send button)Frontend

Add "Renewals Due" widget to main dashboardFrontend

Testing

Unit tests for churn prediction scoring logicTest

E2E: generate recommendation → review → send offer flowTest

▼

💰

1.3 Smart Collections Agent

L 0/0

AI / Logic

Design dunning sequence: Day 1 (friendly SMS) → Day 3 (email) → Day 5 (firm notice) → Day 10 (payment plan offer) → Day 15 (manager escalation)AI/ML

Build tone-adaptation logic: first-time late vs. repeat offender vs. historically excellentAI/ML

Create payment plan generator: auto-calculate installments based on balance and tenant incomeAI/ML

Backend

Create CollectionSequence model (tenant FK, current_step, started_at, last_action, status)Data

Create CollectionAction model (sequence FK, action_type, message_content, sent_at, response)Data

Create Celery task: daily scan for overdue balances, advance sequences, trigger actionsBackend

Add GET /collections/active/ — list active collection sequencesBackend

Add POST /collections/{id}/offer-plan/ — create payment planBackend

Add POST /collections/{id}/escalate/ — flag for manager reviewBackend

Frontend

Design collections dashboard: delinquent tenants list with sequence status, balance, days lateDesign

Build collections timeline component showing sequence steps + tenant responsesFrontend

Add payment plan creation modal (split amount, set dates, preview schedule)Frontend

Add "Delinquent" indicator on unit detail + property overviewFrontend

Add "Collections" widget to main dashboard (count, total balance, escalated)Frontend

Testing

E2E: overdue balance triggers sequence, advances through steps, escalation worksTest

▼

📊

1.3E Collections Intelligence Enhancements

L 0/0

Depends on: 1.3 Smart Collections Agent (built). Research-backed enhancements from industry analysis + AI opportunity study.

Dashboard — P1 Quick Wins

D-1: Enhanced KPI cards — add trend arrows (% vs last month), sparklines (6-month history), benchmark indicators (green/amber/red vs industry avg)Frontend

D-2: Collections Health Score gauge — composite 0-100 score (collection rate 40%, aging trend 30%, delinquency rate 20%, recovery rate 10%). Large circular gauge widget.Frontend

D-6: Enhanced table columns — add risk score badge, last contact date, payment history sparkline (6 dots: green=paid, red=missed), assigned agent. Add column sorting. Add bulk actions (SMS, assign, export).Frontend

D-10: Quick filter chips strip — clickable chips above table: Critical 90+, Payment Plans, Legal Cases, First-Time Late, Chronic Late, My CasesFrontend

Dashboard — P2 Intelligence Widgets

D-3: Collections Funnel widget — visual funnel: All Tenants → Late → Notice Sent → Payment Plan → Legal → Eviction. Conversion rates between stages. Clickable to filter table.Frontend

D-4: AI Action Queue — scrollable top-10 list of AI-recommended actions across all active cases with confidence %, urgency. Quick execute/dismiss/snooze buttons.AI/ML

D-5: Interactive aging chart — click bucket bar to filter table below. Show count + amount labels. Trend indicators per bucket vs last month.Frontend

D-8: Real collections chart data — replace synthetic placeholder data in collection-chart.component with actual monthly billed vs collected from backend API.Backend

Detail Page — P1 Quick Wins

C-1: Payment History Timeline — visual month dots across lease life. Green circle (on-time), amber (late), red (missed), gray (future). Hover shows details. Place above tabs.Frontend

C-2: Upgraded AI Recommendation Panel — show confidence %, reasoning (“3 similar cases responded to SMS in 24hrs”), expected outcome, 2-3 alternative actions with scores. One-click execute per option.AI/ML

C-3: Financial Summary Card — total lease value, total paid, outstanding, security deposit, late fees, net exposure (outstanding - deposit), projected cost if evicted (rent + legal + vacancy).Frontend

C-7: Escalation Workflow Visualization — horizontal pipeline: 5 levels with current highlighted, dates entered, time in level, auto-trigger date for next level.Frontend

Detail Page — P2 Intelligence Features

C-6: Smart Payment Plans dialog — AI generates 3 options (aggressive/standard/lenient) with projected recovery rates. Manager picks one. Replace hardcoded 3-installment logic.AI/ML

C-8: Cost of Inaction Calculator — escalating cost timeline: today, +30d, +60d, +90d showing lost rent + legal fees + vacancy. Daily cost rate. Motivates action.Frontend

C-10: AI-Drafted Communications — pre-populate SMS/email dialogs with AI-drafted messages based on tenant situation, history, tone. Tone selector: Friendly/Firm/Formal/Final Warning.AI/ML

Backend Support

Add GET /collections/stats/trends/ — monthly trends for last 6 months (overdue, collected, rate, new delinquencies)Backend

Add GET /collections/funnel/ — counts per escalation stage with conversion ratesBackend

Add GET /collections/ai-queue/ — top N recommended actions across all active cases with confidence scoresBackend

Add GET /collection-cases/{id}/similar/ — find comparable cases by amount range, days overdue, escalation levelBackend

Add POST /collection-cases/{id}/draft-message/ — AI generates communication draft given channel and tone preferenceBackend

Add POST /collection-cases/{id}/smart-plans/ — generate 3 AI payment plan options with projected recoveryBackend

Extend /collections/stats/ to include health_score, trend data, benchmarksBackend

Testing

E2E: Dashboard loads with health score, funnel, AI queue, enhanced KPI cardsTest

E2E: Detail page shows payment timeline, financial summary, escalation pipeline, upgraded AI panelTest

E2E: Smart payment plan dialog generates 3 options, selecting one creates planTest

▼

Phase 2 — Revenue: Turnover + Pricing + Maintenance

▼

🔧

2.1 Turnover Orchestration Agent

XL 0/0

Depends on: 1.1 Lifecycle State Engine (notice/vacated states)

Backend & Logic

Create TurnoverPipeline model (unit FK, status: notice_given / inspection_scheduled / inspected / make_ready / listed / leased)Data

Create DepositDeduction model (pipeline FK, item description, amount, category, photo_url)Data

Build pipeline orchestrator: auto-create inspection WO when notice given, auto-create turnover WOs after inspectionBackend

Add deposit return calculator: deposit held - deductions = refund amountBackend

Add GET /turnovers/active/ — list active turnover pipelines with statusBackend

Add POST /turnovers/{id}/advance/ — move to next pipeline stageBackend

Auto-generate turnover work orders: painting, carpet clean, appliance check, deep cleanAI/ML

Turnover cost estimator based on tenant tenure and unit condition historyAI/ML

Frontend

Design turnover pipeline Kanban view (columns: Notice → Inspect → Make Ready → Listed → Leased)Design

Build turnover pipeline board componentFrontend

Build deposit deduction form with line items and photo uploadFrontend

Add turnover timeline to unit detail (when in notice/vacant state)Frontend

Testing

E2E: full turnover pipeline flow from notice to new leaseTest

▼

📈

2.2 Revenue Optimization & Dynamic Pricing

XL 0/0

AI / ML

Build market rent estimation model using unit features (beds, baths, sqft, floor, amenities, location)AI/ML

Build vacancy duration predictor: given asking rent, predict days-to-leaseAI/ML

Build dynamic pricing engine: daily adjustment based on days vacant, demand, seasonalityAI/ML

Build rent-to-market gap calculator across entire portfolioAI/ML

Backend

Create MarketRentEstimate model (unit FK, estimated_rent, confidence, comp_count, calculated_at)Data

Create PricingRecommendation model (unit FK, current_asking, recommended, reason, impact_estimate)Data

Celery task: nightly market rent recalculation + pricing recommendations for vacant unitsBackend

Add GET /pricing/portfolio-gaps/ — rent-to-market analysis by propertyBackend

Frontend

Design revenue opportunity dashboard: portfolio rent gap, property breakdown, MTM candidatesDesign

Build rent-vs-market comparison component on unit detailFrontend

Build pricing recommendations list for vacant units (accept/reject/adjust)Frontend

Add revenue leakage widget to main dashboardFrontend

Testing

Unit tests for pricing algorithm edge cases (zero rent, no comps, extreme values)Test

▼

🛠

2.3 Predictive Maintenance Agent

L 0/0

AI / ML

Build maintenance pattern analyzer: frequency by category, building, season, equipment ageAI/ML

Build failure prediction model: probability of HVAC/plumbing/electrical failure per unit per monthAI/ML

Build preventive maintenance scheduler: optimal timing for inspections/replacementsAI/ML

Backend

Create MaintenancePrediction model (unit FK, category, risk_score, predicted_date, recommended_action)Data

Celery task: weekly maintenance risk scan across portfolioBackend

Add GET /maintenance/predictions/ — high-risk units with recommended actionsBackend

Add POST /maintenance/predictions/{id}/create-wo/ — convert prediction to work orderBackend

Frontend

Build predictive maintenance dashboard: risk heatmap by building, upcoming predictions listFrontend

Add risk indicator to unit detail maintenance tabFrontend

Add "Preventive Maintenance" widget to main dashboardFrontend

Testing

Validate prediction accuracy against historical maintenance dataTest

▼

Phase 3 — Intelligence: Vision + Invoices + Chat

▼

📷

3.1 Move-Out Photo Inspection AI

L 0/0

Depends on: 2.1 Turnover Orchestration (inspection stage)

AI / Vision

Design damage classification taxonomy: carpet, wall, appliance, fixture, flooring, door/windowAI/ML

Build vision prompt for Claude/GPT-4V: detect damage from photo, classify severity, estimate repair costAI/ML

Build wear-vs-damage classifier: distinguish normal wear and tear from tenant-caused damageAI/ML

Build deposit deduction letter generator with photo evidence and itemized chargesAI/ML

Backend

Create InspectionPhoto model (pipeline FK, photo_url, room, ai_analysis JSON)Data

Add POST /inspections/{id}/analyze-photos/ — triggers AI analysis on uploaded photosBackend

Add GET /inspections/{id}/deduction-letter/ — generates PDF deposit deduction letterBackend

Frontend

Build photo upload + AI analysis view: grid of photos with damage annotations overlayFrontend

Build deduction review form: accept/reject/adjust AI suggestions per itemFrontend

Build deduction letter preview with download/send optionsFrontend

Testing

Test with sample damage photos: verify classification accuracy and cost estimatesTest

▼

🗃

3.2 Vendor Invoice Processing AI

L 0/0

Build invoice extraction prompt: vendor name, date, line items, amounts, tax, total from PDF/photoAI/ML

Build work order matching: fuzzy match invoice to open WOs by vendor, unit, description, dateAI/ML

Build price anomaly detector: compare invoice amounts to historical averages for similar workAI/ML

Build GL code classifier: auto-assign expense category based on descriptionAI/ML

Backend

Create VendorInvoice model (vendor, amount, status, extracted_data JSON, matched_wo FK)Data

Add POST /invoices/upload/ — upload + trigger AI extractionBackend

Add POST /invoices/{id}/approve/ + /reject/ + /flag/Backend

Frontend

Build invoice upload + extraction review page (side-by-side: original doc + extracted fields)Frontend

Build invoice approval queue with anomaly highlightsFrontend

Testing

Test extraction accuracy on 20+ sample invoices (different formats, vendors)Test

▼

🧠

3.3 Portfolio Intelligence Chat

XL 0/0

AI / NLP

Design query understanding system: parse natural language into structured portfolio queriesAI/ML

Build tool-use integration: give LLM access to portfolio data APIs as toolsAI/ML

Build response formatter: tables, charts, actionable recommendations from raw dataAI/ML

Build anomaly narrator: explain what's unusual and why it mattersAI/ML

Build daily digest generator: summarize portfolio health in natural languageAI/ML

Backend

Create chat session management: conversation history, context window managementBackend

Create data access layer: secure read-only access to all portfolio data for AI queriesBackend

Add POST /chat/query/ + WebSocket for streaming responsesBackend

Add GET /chat/digest/ — daily AI-generated portfolio summaryBackend

Frontend

Design chat interface: sidebar panel or full page, streaming responses, inline data cardsDesign

Build chat component with streaming text, embedded tables, action buttonsFrontend

Build daily digest notification + full digest pageFrontend

Add suggested queries / quick actions in chat interfaceFrontend

Testing

Test 20 sample queries for accuracy, relevance, and response qualityTest

▼

Phase 4 — Delight: Staging + Screening + Abstraction

▼

🏠

4.1 AI Virtual Staging & Listing Generation

L 0/0

Integrate image generation API for virtual staging (empty room → furnished room)AI/ML

Build listing description generator from unit features, amenities, neighborhood dataAI/ML

Build style selector: Modern, Scandinavian, Traditional, Mid-Century staging optionsAI/ML

Backend

Create StagedPhoto model (unit FK, original_url, staged_url, style, status)Data

Add POST /staging/generate/ — upload empty room photos, get staged versionsBackend

Add POST /listings/generate-description/ — AI-written listing copyBackend

Frontend

Build staging interface: upload photos, select style, before/after comparison sliderFrontend

Build listing preview with generated copy + staged photos + publish actionsFrontend

Testing

Quality review staged photos for realism and accuracyTest

▼

👤

4.2 AI Tenant Screening

L 0/0

Build applicant risk scoring model: income-to-rent ratio, credit indicators, rental history, employment stabilityAI/ML

Build recommendation engine: approve / approve with conditions / deny, with reasoningAI/ML

Backend

Create Application model (unit FK, applicant info, income, employment, status, ai_score, ai_recommendation)Data

Add POST /applications/ + POST /applications/{id}/screen/ — trigger AI screeningBackend

Add POST /applications/{id}/approve/ + /deny/ with reason trackingBackend

Frontend

Build application review page: applicant details, AI score breakdown, approve/deny actionsFrontend

Build application intake form (or integrate with online application portal)Frontend

Testing

Fair housing compliance review of screening criteriaTest

▼

📄

4.3 Lease Document Abstraction

M 0/0

Build lease extraction prompt: rent, dates, renewal terms, pet policy, early termination, special clausesAI/ML

Build clause risk classifier: flag non-standard or risky clauses with explanationAI/ML

Backend

Add POST /leases/extract/ — upload PDF, return extracted termsBackend

Add POST /leases/extract/{id}/apply/ — apply extracted data to lease recordBackend

Frontend

Build extraction review page: PDF viewer + extracted terms side-by-side, edit + confirmFrontend

Highlight flagged clauses with risk level and recommendationFrontend

Testing

Test extraction on 10+ lease PDFs of varying formats and lengthsTest

▼

📷

4.4 Move-In Inspection AI

M 0/0

Backend

Create MoveInInspection, MoveInRoom, MoveInItem, MoveInPhoto Django models with migrationsBackend

Build inspection CRUD API: create, list, detail, upload photos, finalizeBackend

Build digital signature endpoints: submit signature, verify, generate reportBackend

Build comparison endpoint: match move-in vs move-out rooms/itemsBackend

Build photo quality check prompt (blurry, dark, angle detection)AI/ML

Build condition analysis prompt (Claude Vision: items, ratings, pre-existing issues)AI/ML

Build comparison prompt (move-in vs move-out, classify as pre-existing/new damage)AI/ML

Frontend

Build room-by-room inspection flow with photo upload gridFrontend

Build condition report view with signature padFrontend

Build side-by-side comparison view (move-in left, move-out right)Frontend

Testing

Test complete inspection flow: create → upload → analyze → sign → compareTest

▼

📜

4.5 AI Lease Generation & E-Signing

L 0/0

Backend

Create LeaseTemplate, LeaseClause, GeneratedLease, LeaseSignature, LeaseAddendum, JurisdictionRule modelsBackend

Build clause library with 50+ categorized clauses across 25 categoriesBackend

Build jurisdiction rules engine (deposit limits, notice periods, required disclosures per state)Backend

Build lease generation API: assemble clauses, substitute variables, generate HTML/PDFBackend

Build e-signature flow: create signing requests, public signing endpoint, audit trailBackend

Build renewal generation: pull current terms, apply escalation, highlight changesBackend

Build lease assembly prompt (clause ordering, variable substitution, compliance check)AI/ML

Build risk assessment prompt (per-clause risk scoring with dispute history)AI/ML

Build translation prompt (multi-language with binding-language disclaimer)AI/ML

Frontend

Build lease generation form (terms input + clause preview + risk panel)Frontend

Build tenant-facing signing page (document viewer + signature pad + audit)Frontend

Testing

Test lease generation for 3 different jurisdictions with jurisdiction-specific clausesTest

▼

🏠

4.6 Tenant Onboarding Automation

L 0/0

Backend

Create OnboardingChecklist, OnboardingTask, OnboardingMessage, KeyHandoff, PropertyKnowledgeBase modelsBackend

Build auto-trigger: lease signing creates checklist with context-aware task generationBackend

Build document collection with AI verification endpointBackend

Build key handoff tracking (physical + digital smart lock codes)Backend

Build scheduled communication sequence (Celery tasks)Backend

Build document verification prompt (validate renter's insurance, utility confirmation, etc.)AI/ML

Build tenant FAQ chatbot with property knowledge baseAI/ML

Build personalized utility guide generatorAI/ML

Frontend

Build onboarding timeline view with task status, blocking items, key trackingFrontend

Build tenant FAQ chatbot interfaceFrontend

Testing

Test complete onboarding flow: trigger → tasks → documents → keys → first weekTest

Phase 5 — Operations

Listing Syndication • Showing Management • Tenant Communications • Vendor Intelligence

▼

📣

5.1 AI Listing Syndication & Marketing

L 0/0

Backend

Create Listing, ListingPlatform, ListingLead, ListingAnalytics, CompetingListing modelsBackend

Build listing publish/delist API with multi-platform syndicationBackend

Build unified lead inbox with AI scoringBackend

Build analytics sync (Celery task, every 6 hours)Backend

Build auto-delist trigger on lease signingBackend

Build platform-specific content optimization promptAI/ML

Build lead qualification scoring promptAI/ML

Build competitor analysis promptAI/ML

Frontend

Build listing dashboard with platform status and analyticsFrontend

Build lead inbox with scoring badges and response actionsFrontend

Testing

Test publish/delist flow across 3+ platformsTest

▼

🚪

5.2 Smart Showing & Tour Management

L 0/0

Backend

Create Prospect, Showing, ShowingFeedback, ShowingSlotConfig, NurtureMessage modelsBackend

Build showing scheduling with pre-screening and identity verificationBackend

Build self-guided tour flow: access code generation, expiration, trackingBackend

Build reminder sequence (24h, 2h, 30min Celery tasks)Backend

Build prospect nurturing engineBackend

Build no-show prediction algorithm (6 weighted factors)AI/ML

Build feedback NLP analysis prompt (sentiment, themes, actionable insights)AI/ML

Frontend

Build showing calendar view with color-coded types and risk badgesFrontend

Build conversion funnel dashboardFrontend

Testing

Test complete showing flow: schedule → remind → complete → feedback → nurtureTest

▼

💬

5.3 Tenant Communications Intelligence Hub

L 0/0

Backend

Create TenantConversation, TenantMessage, CommunicationTemplate, BroadcastMessage, TenantSentimentLog modelsBackend

Build inbound webhook for SMS/WhatsApp/email message ingestionBackend

Build conversation CRUD with reply, close, escalation endpointsBackend

Build broadcast messaging system with recipient filteringBackend

Build auto maintenance work order creation from tenant messagesBackend

Build message classification prompt (category, urgency, intent, sentiment)AI/ML

Build auto-response engine with real tenant data substitutionAI/ML

Build churn prediction from communication patternsAI/ML

Frontend

Build unified inbox with conversation list, thread view, and tenant profileFrontend

Build at-risk tenant dashboard with sentiment chartsFrontend

Testing

Test 5 message classifications + auto-response + emergency escalationTest

▼

🛠

5.4 Vendor Performance Intelligence

L 0/0

Backend

Create VendorScorecard, VendorDispatchRecommendation, WarrantyRecord, VendorContract, RepeatVisitAlert, PriceBenchmark modelsBackend

Build scorecard calculator service (6 weighted metrics)Backend

Build warranty check + smart dispatch recommendation engineBackend

Build price benchmarking and repeat visit detectionBackend

Build contract/insurance/license expiry monitoringBackend

Build dispatch reasoning prompt (vendor specialty matching + scoring)AI/ML

Integrate price analysis into existing invoice extraction pipelineAI/ML

Frontend

Build vendor scorecard dashboard with sparkline trendsFrontend

Build warranty tracker and dispatch recommendation panelFrontend

Build vendor comparison viewFrontend

Testing

Test scorecard calculation + warranty check + price benchmarking + repeat detectionTest

UnitCycle AI — Build Checklist