Utility meter representing AI-driven invoice extraction and contract validation for portfolio management

Automated Utility Contract Intelligence

AI-driven invoice extraction, contract validation, and usage forecasting for utility portfolio management

Results that drive change

faster invoice validation
of invoices automatically verified against contracted rates
supplier invoices processed per month

Managing Utilities oversees utility contracts across a large client portfolio, handling supplier communications, invoice retrieval, rate verification, and usage forecasting. New Gradient built an integrated AI platform combining OCR, large language models, time series forecasting, and robotic process automation within a single web portal, fully automating the contract monitoring and invoice management workflow.

Manual verification across complex, inconsistent invoices

Utility brokers and contract managers spend significant time on repetitive administrative tasks: logging into supplier portals, downloading invoices, cross-referencing billed rates against contracted terms, and flagging discrepancies. The invoices themselves are the core challenge. Utility bills are rarely clean structured data. They combine partial tables, free-text charges, variable layouts across suppliers, and scanned PDFs with inconsistent formatting.

Manual verification against contracted rates is tedious, error-prone, and does not scale as a client portfolio grows. Billing errors and overcharges slip through, and staff time that should go toward client relationships gets absorbed by document processing.

OCR, LLMs, and automated rate validation in a single platform

We built a client-facing web portal that automates the full invoice lifecycle. Automated web scraping retrieves utility documents and scanned PDFs from supplier platforms on schedule, eliminating manual downloads entirely. A specialised OCR pipeline with pre-processing tuned for utility invoice layouts extracts structured data from complex, partially tabular documents that defeat generic extraction tools.

The extracted data feeds into large language models that validate every line item against pre-agreed contract rates, identifying discrepancies and potential billing errors automatically. Rather than relying on rigid rule-based matching that breaks when invoice formats change, the LLM approach handles the variability inherent in real-world utility billing across multiple suppliers and document styles. Over 2,000 supplier invoices now flow through the system every month with every single one verified against contracted terms, replacing selective manual spot-checks with comprehensive coverage.

The platform also integrates time series forecasting models trained on historical consumption data and seasonal patterns, giving Managing Utilities forward visibility on expected usage and spend across their client portfolio. This combines backward-looking verification with forward-looking intelligence in a single interface.

From reactive checking to proactive contract management

Invoice validation that previously consumed hours of staff time per billing cycle now runs 5x faster, freeing the team to focus on client advisory work and contract negotiation rather than document processing. Billing discrepancies that would previously have gone undetected are surfaced automatically, and usage forecasting gives both Managing Utilities and their clients better visibility into future costs before renewal decisions.

The system scales with the portfolio: as Managing Utilities adds clients, invoice volume grows without a proportional increase in manual workload.