When people think of carbon accounting, they often imagine endless spreadsheets and millions of lines of tedious emissions calculations. For the past 30 or so years, this has largely been the reality.
That’s not to say it hasn’t been a crucial process. Measuring an organisation’s carbon footprint is essential to understanding its environmental impact, identifying opportunities for emissions reduction, and tracking progress over time. Today, with increasing urgency to limit global temperature rise within acceptable thresholds, carbon accounting has never been more important. However, despite its significance, it has never been the most glamorous or exciting task.
With the rapid advancements in artificial intelligence (AI) and natural language processing (NLP) models like ChatGPT, there’s growing speculation about whether AI can fundamentally change the nature of carbon accounting. Over the past few years, these technologies have revolutionised manual, time-consuming tasks – automating email drafting, extracting key insights from dense reports, and debugging code with unprecedented efficiency. Given the highly manual nature of carbon accounting, it seems like the perfect candidate for AI-driven automation.
But is it really possible? In this blog, we’ll explore AI’s potential to transform carbon accounting and the extent to which it can replace human expertise.
How AI Can Improve Carbon Accounting
Carbon accounting has long been a highly manual and tedious process, making it ripe for AI-driven improvements in efficiency and accuracy. This is an overwhelmingly positive development – by reducing the burden of emissions measurement, AI could encourage more organisations to track their carbon footprints while freeing up sustainability professionals to focus on the strategic aspects of decarbonisation.
Here are some key areas where AI can enhance carbon accounting:
- Classification of Data: Under the Greenhouse Gas (GHG) Protocol, all activity data must be classified with an appropriate emissions factor – a coefficient representing the average rate of CO2e per unit of activity. This is especially tedious in spend-based carbon accounting, where procurement ledgers often contain hundreds of thousands of lines. AI can automate this process by utilising NLP to rapidly match activities to the most relevant emissions factors, reducing both time and error.
- Data Structuring and Cleaning: One of the biggest challenges in carbon accounting is dealing with messy, inconsistent data. AI-powered tools can structure and clean datasets, ensuring that emissions calculations are based on well-organised inputs. This minimises the time spent manually consolidating spreadsheets and streamlines the entire accounting process.
- Accuracy and Error Detection: AI can detect anomalies and errors far more efficiently than human analysts. By cross-referencing data with historical patterns and known emissions benchmarks, AI can flag inconsistencies and suggest corrections, helping organisations generate more reliable emissions estimates.
Taken together, these capabilities suggest that AI, even with minimal user intervention, can generate a fairly accurate carbon footprint from raw data. But is “fairly accurate” good enough?
The Limits of AI: Where Human Expertise is Essential
While AI has the potential to automate much of the carbon accounting process, it will never fully replace human expertise. The main reason? Carbon accounting is not an exact science. Unlike financial accounting, which operates under standardised frameworks, there are no universal, objective rules that define correct or incorrect emissions calculations.
This lack of rigid standards is intentional. Emissions measurement is inherently complex – organisations cannot physically monitor every emissions source across their value chains. Instead, the GHG Protocol relies on estimation methodologies based on available activity data, such as miles driven in a company vehicle or dollars spent on materials.
These estimation methods serve an important purpose: they make it easier for organisations to start measuring their impact without requiring highly complex, academic studies. However, as net-zero deadlines approach, the need for greater accuracy is becoming increasingly critical. We can no longer rely on “good enough” estimates – we need to transition toward more complex, real-world assessments of emissions sources that facilitate meaningful decarbonisation.
Why AI Alone Falls Short
Even the most advanced AI models will be insufficient to handle several fundamental aspects of carbon accounting. Some of these include:
Contextual Understanding & Organisational Complexity
Every organisation has a unique operational model and supply chain structure that influences its carbon footprint. AI may generate emissions estimates based on standard methodologies, but it cannot fully grasp the nuances of business decisions – such as shifting production locations, renegotiating supplier contracts, or optimising logistics – that impact emissions in non-obvious ways.
Data Gaps & Modelling
AI models depend on structured, high-quality data, but carbon accounting often involves working with incomplete or inconsistent information. Scope 3 emissions, in particular, are challenging due to the lack of primary data and reliable supplier disclosures. As such, gap-filling exercises utilising advanced modelling techniques are often required. Human judgment is essential for making informed assumptions that feed into these models and ensure they operate at the highest level of accuracy.
Strategic Decision-Making & Policy Alignment
AI can identify trends and suggest reduction strategies, but it cannot determine whether these strategies are feasible within a company’s financial, operational, and regulatory constraints. Human experts are required to align AI-driven insights with corporate goals, industry regulations, and evolving climate policies.
Regulatory Compliance
Navigating the legal and compliance risks associated with emerging sustainability regulations is a critical aspect of carbon accounting. AI can assist in tracking regulatory changes, but it lacks the expertise to interpret complex legislation and apply it to an organisation’s specific circumstances. Human oversight is necessary to ensure compliance with both national and international climate policies.
Ethical and Social Justice
Decarbonisation is not just a technical challenge – it involves social and ethical dimensions, such as supply chain labour practices and environmental justice. AI cannot fully account for these broader impacts, nor can it ensure that sustainability initiatives align with ethical best practices.
The Need for Human Oversight & Continuous Learning
AI models are only as good as the data they are trained on. They may struggle with novel challenges or emerging trends that fall outside historical patterns. Human oversight is critical to refining methodologies, validating AI-generated insights, and adapting carbon accounting practices as new regulations and technologies emerge.
The Future of AI in Carbon Accounting: A Hybrid Approach
AI has already begun to transform carbon accounting by automating data-intensive tasks, improving accuracy, and reducing the time burden of emissions measurement. However, it is not a standalone solution. The complexity of emissions measurement, decarbonisation strategy, and regulatory alignment requires human expertise to interpret AI-generated outputs and apply them in a meaningful way.
At City Science, we have already begun to use basic NLP models to streamline our classification processes and improve outcomes for clients. However, we – just as any organisation integrating AI into its core processes – are still seeking the optimal balance between automation and human intervention and oversight.
To this end, the most effective approach is a hybrid one: AI assists sustainability professionals by handling routine calculations, structuring data, and identifying potential errors, while human experts provide the strategic, contextual, and ethical considerations necessary to drive real decarbonisation. Rather than replacing carbon accountants, AI will serve as an invaluable tool – enhancing efficiency, improving data quality, and enabling deeper insights that accelerate the transition to net zero.
Want to learn more about how we at City Science are using AI to help our clients measure their environmental impact and develop strategies to reduce emissions across Scopes 1, 2, and 3? Reach out to us at info@cityscience.com.