Skip to main content
A view looking up from the ground surrounded by skyscrapers

Artificial intelligence (AI) is estimated to add $2.6 billion to $4.4 billion to the global GDP annually (and we don’t just mean ChatGPT, which you may have heard of). The technology’s impact on productivity and efficiency across industries is leading the way for new economic opportunities. Interestingly, carbon accounting is another area where leveraging AI can be incredibly useful.

Problem: One of the biggest challenges in fighting climate change is inaccurate and low-quality emissions data. With unreliable carbon emissions data, efforts to reduce emissions can go in vain as they might not make the impact needed to stop global warming in its tracks.

Solution: AI can help change that and allow more accurate target setting for emissions reduction, and by extension, help countries reach their climate goals. If used correctly, AI can empower the world to prevent global temperatures from rising beyond 1.5 °C by 2050.

Let’s dive in!

Current carbon accounting challenges

Carbon accounting, which involves measuring greenhouse gas (GHG) emissions, is critical to the success of climate efforts. Unfortunately, in many cases, there are glaring inaccuracies in carbon accounting. Various reasons can be attributed to this issue, from the unavailability of data to the lack of oversight.

A 2021 Washington Post report found some countries were using flawed data and underreporting emissions data. The investigation found measurable gaps between the emissions reported and those generated – less than ideal, but likely due to challenges that may sound familiar.

The three main challenges often encountered with carbon accounting are:

Scope 3 emissions

It’s estimated that Scope 3 emissions account for up to three-fourths of a company’s emissions. Unfortunately, many companies use estimates for Scope 3 or value chain emissions, which, more often than not, end up being underreported.

Measuring Scope 3 emissions is understandably challenging, but it’s also instrumental for companies to tackle their actual carbon footprint. Many companies don’t even include Scope 3 emissions in their ESG reporting.

Poor quality data

Even when data on emissions, particularly Scope 3 emissions, is available, it’s poor quality. For instance, many companies resort to using spend-based data.

Spend-based data has many fundamental issues, as it’s inaccurate to measure emissions. In comparison, activity-based data is more accurate. However, gathering activity-based data requires more research and resources – something many companies can’t afford.

Similarly, in many instances where Scope 3 emissions make it into the ESG reporting, they’re based on reported sector intensity averages, which are ridiculously unreliable.

Estimates vs. real data

When data is unavailable, many companies resort to estimates to meet their reporting requirements. However, estimates of value chain emissions are mainly inaccurate. Think of estimates as the villain of good carbon accounting practices.

In contrast, primary data from companies in the value chain is accurate, which can help companies improve their emissions accounting.

A quote from the article with a black background

Ways AI can improve carbon accounting

AI’s ability to process and learn from existing data can be leveraged to improve emissions estimates wherever used in carbon accounting. Furthermore, with machine learning (ML), organisations can model, measure, and reduce emissions.

Here are two ways AI can be used for measuring emissions more accurately:

Advanced data analytics

Large enterprises with a global supply chain find measuring their value chain emissions challenging. As they get data from various sources and in different formats, structuring and analysing it can be a laborious task.

Using powerful algorithms, companies can integrate and analyse data from all their suppliers and measure emissions independently. In other words, AI can help consolidate and analyse data on Scope 3 emissions. Companies can use this data to set more realistic targets and develop ways to reduce value chain emissions.

Compared with traditional data analytics, AI-powered analytical tools may provide deeper insights into the data – especially the terabytes (or yottabytes – Google that one) of unstructured data you thought you’d never use.
Companies and governments worldwide can use ML to make sense of what would otherwise be random climate data from the value chain. AI models can be trained to provide navigable and actionable strategies aligned with global climate goals.

Predictive modelling

AI can also be used in the carbon accounting of future activities. As time is of the essence, governments and organisations worldwide must determine the environmental impact of future projects/products.

AI can create predictive models based on historical data and carbon emissions associated with various activities. Using the findings from existing data, AI models can quickly predict how much emissions a particular project will emit.

Using predictive modelling for estimating emissions for any future projects can help companies make the project sustainable from the get-go. Instead of completing a project and then measuring its impact, AI can be used to calculate emissions before it’s even begun. Magic.

Addressing the challenges of using AI for carbon accounting

Before we get too carried away with AI’s abilities to improve carbon accounting and reporting, we should also look at the pitfalls (and we don’t mean Skynet). Here are two worth considering:

Data bias

As AI models are trained on data, bias can find its way into the algorithms. Like humans, even AI can be biased, as seen in multiple cases of recruitment software.

Before any company opts to use AI models for carbon accounting, removing any biases from the data the model is training on is essential. This may require effort and investment to ensure the data is as accurate as possible.

AI computing’s carbon footprint

When discussing anything related to AI and climate change, we need to address the elephant in the room – AI’s own impact on climate. The hardware resources (e.g. servers) needed to run an AI or ML model consume a ton of power. Put simply, AI computing comes with a high energy consumption.

According to one study, training a single AI model can produce emissions equivalent to 41 roundtrip flights from New York to Sydney.

Fortunately, efforts to make AI research more energy-efficient and climate-friendly have already begun.

A skyline shot of Dubai with skyscrapers poking through clouds

Taking a balanced approach

There’s no doubt that AI can be a powerful tool, and we’re lucky enough to witness the adoption of it in innovative ways – including in the battle to stop climate change. AI’s powerful learning abilities can be used to analyse climate data, predict potential emissions, and find creative solutions to complex climate issues.

Simultaneously, a cautious approach must be taken, ensuring AI’s estimates and recommendations help achieve the desired results. The full extent of AI’s impact on carbon accounting and reporting is yet to be seen, but the future looks promising and we’re excited to see how it unfolds.

If you’re looking to calculate carbon emissions for buildings, comundo is the right solution for you. Using proprietary technology, comundo gets hold of real data, not estimates. You can see the real carbon footprint of properties in your portfolio and take steps to make them more energy-efficient.

Get accurate emission data today.

Base your optimisation and reporting on facts – not estimates. Join comundo.

Ryan Stevens

Technical content creator

Ryan is a senior technical content creator, helping tech businesses plan, launch, and run a successful content strategy. After an extensive academic career in engineering, he worked with dozens of tech startups and established brands to reach new clients through proven content creation strategies.