CarbonTrail cuts generative AI costs by 88% for sustainable emissions intelligence

CarbonTrail’s mission is to help businesses measure and reduce emissions, with a goal of avoiding 1 billion tons of CO₂ by 2050. However, traditional methods—manual processes and broad spend-based estimates—were too slow, inaccurate, and resource-heavy to scale. Banks are further challenged by regulatory requirements that demand processing hundreds of thousands of customer records across multiple systems, a scale that manual approaches could never achieve. This was the case at the Bank of New Zealand (BNZ), which set a target to measure the emissions of 50 percent of its SME customers.

CarbonTrail sought to build a platform capable of ingesting large volumes of accounting and invoice data, delivering precise insights, minimizing the emissions of AI workloads, and meeting strict data sovereignty rules with in-region hosting. As founder Tom Hallam notes, “Banks are under pressure to understand their financed emissions—the greenhouse gases tied to lending and investments, which represent their largest climate impact. Achieving that with speed and precision required an AI-powered platform that could deliver scale, accuracy, and trust.”

To meet these demands, CarbonTrail built an AI-powered emissions measurement platform on AWS. At its core, the platform runs a document-analysis pipeline on Amazon Bedrock, processing unstructured data such as invoices and receipts to extract the details required for accurate emissions calculations. To increase efficiency, CarbonTrail runs workloads on AWS Inferentia—part of AWS’s custom silicon portfolio—achieving higher throughput with lower emissions per token. The team integrates generative AI models such as Llama through Amazon Bedrock foundation models, enabling sustainable carbon accounting at scale. Finally, AWS Fargate and AWS Lambda provide containerized, serverless applications that give the platform elastic operations and seamless scalability.

Hallam says, “As a team, we’re specifically looking for tools and services and hardware that allow us to deliver insights, but at a fraction of the power usage or a fraction of the emissions associated with a standard generative AI approach. That’s why we landed on AWS Inferentia.”

The platform is hosted in-region to meet sovereignty requirements and supports deployment models such as Amazon Virtual Private Cloud (Amazon VPC) peering for banking clients. Hallam adds: “Using AWS lets us run models entirely within our environment, so data stays contained and compliant with our clients’ security expectations. This isn’t a one-size-fits-all SaaS product—it’s designed to integrate directly into a client’s environment and deliver secure, sustainable insights at scale.”

Building on this foundation, CarbonTrail also introduced CarbonAPI—a developer-facing service that extends the same invoice-level emissions measurement to partners and fintechs. Powered by AWS Inferentia, the API enables organizations to embed standardized, audit-ready emissions insights directly into their own applications and services, helping scale sustainability reporting far beyond individual clients.

Hallam says, “We can handle hundreds of thousands of customer records without extra overhead. Using AWS Inferentia, we’re also cutting emissions from our AI workloads—essential to our mission.”

By running its emissions measurement platform on AWS, CarbonTrail delivers performance and efficiency gains—achieving 87% reduction in processing time and 88% lower cost than comparable GPT-4 with embeddings. The platform also reduces low-confidence classifications by up to 40 percent and keeps data in-region to support sovereignty and compliance.

Through CarbonAPI, banks, fintechs, and enterprises gain the ability to process transaction and invoice data at volumes that would be impractical to manage manually, while also accelerating time-to-market. The ready-to-use API replaces months of development with emissions tracking that can be integrated in minutes, making it easier for organizations to embed sustainability insights directly into their products and services.

 

The impact is already clear. Alex West, head of sustainable finance, growth sectors at BNZ, notes: “CarbonTrail’s CarbonAPI on AWS is helping us progress toward our goal of measuring emissions across our SME customers. Accurate, invoice-level insights delivered in-region give us the confidence that we’re supporting customers to meet their sustainability commitments, as well as the bank’s ambitions.”

Looking ahead, CarbonTrail plans to expand the reach of its CarbonAPI while continuing to refine its AI models to further optimize the emissions footprint of inference and automate operational workflows.