Financial administrators managing foundations and endowments operate across a fragmented landscape of spreadsheets, bank PDFs, and disconnected accounting systems. A Swedish financial services organization was spending days every month extracting data from multiple sources, manually reconciling numbers, and assembling reports. They needed a single platform that could talk to all of their systems, unify the data, and generate accurate reports in minutes.
Problem
The organization manages multiple foundations with separate bookkeeping requirements. Financial data was scattered across Visma (their ERP), bank statements (PDF extracts), Excel spreadsheets, and SIE files (Swedish accounting format). Quarterly report generation required manual data extraction, transformation, and reconciliation—a process that consumed 40+ hours monthly and was error-prone.
No single system owned the truth. Reconciling Visma balances against bank statements and regulatory reports meant jumping between tools, hunting for discrepancies, and making manual adjustments. Board-level visibility into multi-entity financial health was nearly impossible without this manual assembly work.
Solution architecture
API orchestration across systems
The platform connects to Visma via OAuth 2.0, automatically fetching live account balances and transaction data. Concurrently, it processes PDF bank statements and fund reports using Google Document AI for intelligent table extraction. Excel templates and SIE files are parsed and normalized.
Intelligent reconciliation engine
A two-pass reconciliation algorithm first attempts exact matching on amount and date across sources. Unmatched transactions are then scored using fuzzy logic—comparing descriptions, amounts (within tolerance), and date proximity. Confidence scores (0–100) let administrators flag uncertain matches for review.
Vector-indexed document processing
Document chunks from PDFs are embedded and indexed, enabling semantic search across fund statements, regulatory reports, and historical documents. Classification rules stored in a Firestore database categorize transactions (interest, dividends, realized gains) automatically.
Reconciliation time
From days to minutes
Data sources unified
4+ systems
Report generation time
< 5 minutes
Reconciliation accuracy
99.8%
Real-time visibility
Live connections to Visma and fund managers provide board-level dashboards showing multi-entity financial health, investment performance, and regulatory compliance status. No stale data; no manual exports.
Insight
System independence
The platform's abstraction layer isolates business logic from system-specific APIs, making it trivial to swap vendors or add new data sources without rewriting core reconciliation and reporting engines.
Outcome
Report generation moved from a three-day process to five minutes. Foundation managers now have real-time visibility into account balances, fund valuations, and regulatory compliance metrics across all entities. Audit trails are comprehensive—every number is traceable back to its source system. The organization eliminated manual spreadsheet assembly entirely, reducing errors and freeing staff to focus on strategic financial management rather than data wrangling.
