Daugiabučių Komunalinių PDF Sąskaitų Automatizavimas (Ignitis, ESO, Vilniaus vandenys) ir Nuomininkų Skolų Knyga

1. Kodėl Ignitis, ESO ir Vilniaus vandenų PDF sąskaitų suvedimas į Excel žlugdo portfelio maržą

Daugiabučių nuomos fondai ir co-living operatoriai Vilniuje, Kaune bei Klaipėdoje, valdantys nuo 100 iki 2 500 butų, kiekvieną mėnesį praranda vidutiniškai 27,5 darbo valandos ir 1 120 € vien dėl rankinio Ignitis, ESO, Vilniaus vandenų ir Kauno energijos PDF sąskaitų suvedimo į Excel lenteles. Across a typical 100-door residential building, monthly utility ingestion requires reconciling 300 to 500 separate PDF invoices spanning electricity, cold/hot water submetering, district heating, gas, and communal service charges. When property managers rely on shared spreadsheets and manual email attachments, four systemic revenue leaks compound simultaneously:

• Unassigned Utility PDF Backlog (8%–14% Error Rate): Bulk municipal water and district heating bills from Ignitis, ESO, Vilniaus vandenys, Kauno energija ir Miesto gijos frequently omit the apartment door number on page 1, listing only the physical meter serial number (extracted_meter_serial) or customer contract ID buried on page 3. Manual clerks misfile or delay these invoices past the tenant billing cutoff.
• Broken Mid-Month Tenancy Turnover Proration: When Tenant A moves out on day 14 of a 31-day billing cycle and Tenant B moves in on day 16, standard spreadsheets fail to split the incoming invoice by exact occupancy overlap days (overlap_days / total_period_days) while automatically isolating the 1-day landlord vacancy gap.
• Lost Continuity Between Tenant Arrears and Overpayment Credits: When a tenant rounds up a bank transfer (for example paying 150.00 instead of 120.00), spreadsheets lose track of the -30.00 credit carry-forward (running_balance_cents < 0), triggering disputes on the following month's statement.
• Silent Vacant-Unit Leaks & Consumption Spikes: Without automated trailing 6-month baseline auditing, a running toilet valve or active radiator in a vacant apartment silently burns 350 €/mėn. before anyone inspects the physical meter.

2. 3 Etapų AI Priskyrimo Architektūra (98/100 Automatinis Buto Atpažinimas + 2/100 Išimčių Eilė)

Billaily eliminates manual data entry by combining a multimodal Gemini 2.5 Pro Vision extraction pipeline with a deterministic 3-Pass Auto-Assignment Engine backed by PostgreSQL 16 Row-Level Security (RLS). Every incoming PDF from Ignitis, ESO, Vilniaus vandenys, Kauno energija ir Miesto gijos is parsed into strict JSON coordinates and evaluated through three sequential matching passes:

[Inbound Utility PDF Stream: Ignitis, ESO, Vilniaus vandenys, Kauno energija ir Miesto gijos]
       │
       ▼
[Multimodal Vision Parser (Gemini 2.5 Pro)]
  ├─► extracted_meter_serial: "MTR-884920-B"
  ├─► extracted_address_raw:  "Oak Residence, Apt 14"
  ├─► billing_period:         2026-10-01 .. 2026-10-31 (31 days)
  └─► total_amount_cents:     15500 (€)
       │
       ├─► PASS 1: Exact Meter Serial / Contract Lookup (unit_meters)
       │     └─► Match Found? ──► AUTO_ASSIGNED (confidence = 1.000) [84% of bills]
       │
       ├─► PASS 2: Normalized Street + Door Number Fuzzy Match (units)
       │     └─► Match Found? ──► AUTO_ASSIGNED (confidence = 0.950) [14% of bills]
       │
       └─► PASS 3: Constrained Missing-Utility Candidate Filter (2% Exception Queue)
             └─► SQL Filter: Select ONLY apartments in building X still missing
                 utility_type = 'WATER' for period 2026-10-01..2026-10-31
             └─► 1-Click Manager Assignment ──► UPSERT into unit_meters
                 (Future bills from this meter auto-assign in Pass 1 @ 1.000)

The architectural breakthrough lies in Pass 3 (The 2/100 Constrained Exception Queue). Instead of forcing a property manager to scroll through a dropdown of all 200 apartments when a PDF lacks a door number, Billaily queries GET /api/bills/:id/candidate-units to filter the dropdown exclusively to the 2 or 3 doors in that building that have not yet received that utility type for that billing period. The moment the manager clicks to assign the bill, Billaily writes the extracted meter serial into unit_meters so every subsequent monthly invoice from that meter matches automatically at 1.000 confidence.

3. Finansinė ROI Matematika ir Portfelio Ekonomika Lietuvoje (€ EUR)

For an institutional landlord, co-living operator, or property management firm operating across Rivilė GAMA, B1.lt, Directo, Finvalda ir VMI i.SAF, automating utility PDF ingestion produces immediate, auditable unit-economics returns. Consider a baseline portfolio of 100 apartments processing 500 utility PDFs per month:

• Manual Data-Entry Labor Eliminated: Processing 500 utility PDFs manually (opening email attachments, locating meter serials, typing amounts into Excel, and computing tenant splits) requires 3.3 minutes per bill = 27.5 hours/month. At a loaded property accountant rate of 28 €/val., manual administration costs 770 €/mėn. every single month.
• Unbilled Utility & Vacant-Unit Leak Recovery: Portfolios running manual spreadsheets fail to bill back an average of 2.8% of total utility spend due to orphaned PDFs, mid-month move-out gaps, and undetected vacant-unit water/heating leaks, costing an additional 350 €/mėn. per 100 doors.
• Net Portfolio ROI & Payback Formula: Net Monthly Savings = (Manual_Hours × Hourly_Rate) + Leak_Recovery − Billaily_Pro_Subscription. With Billaily Portfolio Pro priced at €79/month (up to 150 doors), a 100-door portfolio recovers 13 440 €/metus annually with a full payback period of 17 dienų.

4. Architektūrinis Palyginimas: Excel Lentelės vs. Senos Apskaitos Sistemos vs. Billaily 98/100 AI

The table below benchmarks manual spreadsheet workflows against legacy property management suites and Billaily's 98/100 AI Auto-Assignment & Continuous Tenant Ledger architecture:

5. Sudėtingi Atvejai: Nuomininkų Kaita Mėnesio Viduryje, Permokos ir Tuščių Butų Nuotėkiai

Standard OCR wrappers and generic accounting tools break down when confronted with real-world residential tenancy edge cases. Billaily's database engine handles four mission-critical failure modes natively:

1. Mid-Month Tenant Move-In / Move-Out Proration: Suppose a district heating invoice from Ignitis, ESO, Vilniaus vandenys, Kauno energija ir Miesto gijos covers 2026-10-01 to 2026-10-31 (31 days) for €155.00 (15500 cents). Tenant A moves out on October 14 (14 days = €70.00), the unit sits vacant on October 15 (1 day = €5.00 absorbed by the landlord vacancy ledger), and Tenant B moves in on October 16 (16 days = €80.00). Billaily automatically generates separate PRORATED_UTILITY_CHARGE entries with exact proration_days and total_period_days audit trails.
2. Bulk Municipal Invoices Missing Apartment Numbers: Water and heating suppliers frequently issue pdf statements containing only a technical point-of-delivery code (POD / PPE / Zählernummer). Once assigned once via the 2/100 Constrained Queue, the meter serial is permanently bound to the apartment in unit_meters.
3. Continuous Overpayment Credit Carry-Forward: When a tenant pays 150.00 against a 120.00 statement via their Magic Link (/t/[magic_link_token]), Billaily records a running_balance_cents = -3000 surplus credit and automatically deducts 30.00 from next month's utility charge.
4. Silent Vacant-Unit Consumption & 1.5x Variance Spikes: Any utility bill arriving for a unit with zero active tenancies triggers anomaly_flag = 'VACANT_UNIT_USAGE', while any bill exceeding 1.5x the unit's trailing 6-month average triggers HIGH_VARIANCE_SPIKE before the landlord pays the supplier.

7. 4 Savaičių Portfelio Diegimo Planas ir Integracija su Rivilė GAMA / B1.lt / SEPA

Transitioning a 100 to 10,000+ door portfolio from spreadsheets to Billaily requires zero hardware installation and follows a structured 4-week rollout:

• Week 1 — Building & Door Topology Import + Dedicated Inbound Inbox: Import your apartment doors via CSV or create them directly inside /dashboard/reconcile, and activate your dedicated inbound forwarding address (e.g. portfolio@in.billaily.com).
• Week 2 — Historical Utility PDF Drop & 98% Meter Serial Learning: Drag and drop 1 to 3 months of past invoices from Ignitis, ESO, Vilniaus vandenys, Kauno energija ir Miesto gijos. Pass 1 and Pass 2 auto-assign 98% of bills while Pass 3 binds the remaining 2% into unit_meters.
• Week 3 — Continuous Tenancy Ledger & Zero-App-Store Magic Links: Enable automated web statements (/t/[magic_link_token]) where tenants inspect the original provider PDF, verify their daily proration math, and report payments in one click.
• Week 4 — Automated Anomaly Auditing & ERP/SEPA Sync: Activate automatic HIGH_VARIANCE_SPIKE and VACANT_UNIT_USAGE alerts, and optionally connect custom SEPA XML / OpenBanking / Rivilė GAMA, B1.lt, Directo, Finvalda ir VMI i.SAF exports.