SustainLedger: Practical Carbon Reporting, Powered by Smart Processing

Why Scope 3 Is Hard (and Where Tech Helps)

Scope 3 emissions – the indirect emissions from a company’s supply chain and operations (DEFRA guidance) – are notoriously messy. Unlike Scope 1 and 2, the data is fragmented, inconsistent, and often requires tedious manual mapping.

To make this feasible for accountants, SustainLedger combines multiple layers of processing:

1. Basic Lookups

Transactions are first matched against authoritative emissions factor datasets (DEFRA and others). This handles the majority of common business expenses – utilities, fuel, travel, and standard suppliers.

2. Local AI Enrichment

For ambiguous or unusual transaction descriptions, a local embedding model clusters and classifies them to suggest the most likely category. This runs entirely on-premises, so sensitive financial data never leaves the server.

3. Optional Remote AI

In rare cases, we enrich transaction names using remote LLM calls (e.g., OpenAI), but crucially, we never transmit sensitive amounts or client data – only transaction descriptions are sent, and only when the local model can’t confidently classify them.

The Processing Pipeline

From a workflow perspective, the platform is designed for speed and usability. Here’s how data flows through the system:

SustainLedger Processing Pipeline

The goal: turn raw data into a polished carbon report in under 10 minutes, even for relatively large datasets.

Why This Matters

Accountants increasingly need to provide ESG insights alongside financial reporting. SustainLedger makes this practical:

• Fast – Reproducible reports without manual wrangling
• Transparent – Calculations are fully documented, so numbers can be confidently explained to clients
• Smart – Processing gets better over time, reducing effort and errors
• Compliant – PPN 006-ready reports for UK government tenders

“The challenge isn’t just calculating emissions – it’s doing it in a way that’s defensible, repeatable, and fast enough to be practical for busy accounting practices.”

Privacy and Security First

One of the key design decisions was to make the system stateless:

• Transaction data is processed and immediately discarded
• No persistent storage of sensitive financial information
• GDPR-compliant by design
• Local AI processing means most data never leaves the server

This isn’t just good for compliance – it also means accountants can confidently use the platform for multiple clients without worrying about data mixing or retention.

Real-World Results

Early testing shows the system can process typical SME transaction files (500-2000 transactions) in under 5 minutes, with both classification accuracy and performance improving as the local cache grows.

Looking Ahead

The MVP is live at sustainledger.co.uk. Future improvements will include:

• Cross-company trend analysis – Spot patterns across multiple clients
• Benchmarking – Compare against sector averages
• Historical insights – Track emissions reduction over time
• Multi-client dashboards – For accounting practices managing multiple clients

Conclusion

For accountants looking to provide practical, lightweight carbon reporting, SustainLedger shows how smart processing and automation can make what was once tedious and error-prone into something fast, reliable, and insightful.

The platform demonstrates that you don’t need massive infrastructure or expensive consultants to deliver quality carbon reporting – just thoughtful design, local AI where it helps, and a focus on the practical needs of accountants and their clients.