AI tools for ESG reporting: what actually works in practice

March 26, 2026

AI tools for ESG reporting are developing fast, but the gap between the hype and what actually works in practice is still large. Over the last few months I’ve been testing different AI tools and workflows for real sustainability reporting tasks to see what actually saves time and what doesn’t. This is a practitioner’s honest view of where AI really helps, where it falls short, and what’s worth your time.

Where AI actually adds value in sustainability reporting workflows

Wording data requests or making a data collection template:

If you’re the person in your organization responsible for collecting data for your GHG footprint, AI can help you draft the requests in an accurate and useful way. For example, if you’re asking for business travel data and don’t have a template, you can get a bit stuck trying to figure out what data points to ask for and how to make the ask clear for a non-sustainability specialist. With a good prompt that references the GHG Protocol, you can have a clear and concise email in a few seconds. Or even a simple template for your data owners to fill in. Easy half hour saved.

Data collation:

You can use AI to pull data together from multiple sources, so you don’t have to do it by hand. For example, I had 12 separate monthly headcount reports. I needed to look up each office’s headcount for each month, then average it across the year so I have the annual average (to smooth out any annual hiring variations – a best practice!). Instead of opening all 12 PDFs and spending time looking up figures, possibly mis-typing something or pulling from the wrong line accidentally, I gave AI a simple short prompt and loaded in my 12 documents. And it did a perfect job. I then did a manual spot-check to ensure there weren’t any errors. This gave me confidence that the data was correct. Probably saved me about an hour.

Assessing your top suppliers against SBTi target status:

I cannot believe that just a few years ago I used to do this manually. SBTi target dashboard provides a spreadsheet of all the data by company, and you can ask AI to cross-reference your list of suppliers against the SBTi spreadsheet, and make a table. It can also calculate for you what % of your spend is with SBTi-commited suppliers, etc. Huge time saver.

Invoice reading:

Capturing utility data is now super easy. You can load PDFs of your invoices into your AI application and it can pull the data that you need into a table or text, ready for you to copy/paste and send to whoever needs it, or enter it into your data management system. Having a full year worth of utility invoices is no longer a problem! 1 hour saved.

Turning unstructured data into useable insights:

In my world unstructured data is usually sitting in emails. Analyzing information from emails, like client ESG information requests, used to be a manual job where I would transfer information from an email into a spreadsheet. It was a boring task, and usually was given up on after a few weeks once things got too busy. Now AI can analyze this for you, you just need to feed it the emails. But ensure you redact any sensitive information so it doesn’t end up where it shouldn’t, and sense check the outputs, because you just never know. Sanity saved.

Developing internal data management process documents:

Having a documented and clear data collection and management process is increasingly important, and in the past it would take me weeks to develop a decent draft of a data protocol document. Now I can create a decent template within a few minutes, then populate it with my specific information. Many days saved, and a huge help for the assurance process.

Making a first draft of a disclosure:

You can ask AI to pull together a first draft or an outline of a disclosure. Just give it the standard you need to comply with, existing data and narrative you may already have, and have it shape it into something you can then take and tweak.

Ensuring consistency across different reports and comms:

Drafting narrative for multiple reports throughout the year, with varying levels of detail, can easily result in saying something inconsistently and raising red flags. Loading your disclosures into AI and asking it to pick up any inconsistencies will save you lots of time.

Polishing a final draft:

Many of us know the fear and panic that happens in the final few weeks before a report is finalised. Did you forget to update any numbers? Are typos still hiding somewhere? Is the tone of any particular section still off? AI can definitely help with all of this, and in a much shorter timeframe.

Gap assessment against a regulation or a framework:

Load your report in, and the standard you’re aiming to comply with – and ask AI to identify any gaps. So much simpler than before.

AI tools worth knowing about for ESG reporting

General purpose (Claude, ChatGPT):

These work very well, as long as you have enough subject-matter expertise to know when the output is flawed, and you have sorted out any data privacy concerns. I find that Claude, especially, is great at producing templates and spreadsheets, that are actually nicely formatted. I had it create an Excel carbon calculator once, and it delivered after a few iterations.

Custom GPTs/Copilots (Sprig UK, Sprig EU, Muuvment IQ):

These are ChatGPT or Copilot-based, but with custom system prompts and a specific knowledge base. Sprig AI is super useful if you’re in the UK or EU, and want some guardrails that ensure the AI has good context for what you’re likely working on. It’s also completely free. The downside is that you can’t have a private instance of Sprig, so you wouldn’t want to use it to analyze any sensitive data. We wrote a review of Sprig AI here. We haven’t personally tried Muuvment IQ yet, but looks like it allows you to store your own knowledge base, which is super useful.

Specialist AI tools (Briink, EcoAppraise):

Briink helps you fill in questionnaires and requests with AI, while EcoAppraise checks your text against greenwashing risk. We haven’t personally tried many of these types of tools, as there just hasn’t been a strong need. However, depending on your specific circumstances, a specialist tool may be completely worth it. The market is moving very fast in this realm, so it’s a good idea to keep an eye on emerging AI sustainability tech and consider what may be worth trying out.

Where AI still falls short

GHG calculations:

You can definitely do some simple GHG footprint calculations with AI, but I wouldn’t rely on it for anything official or extensive. It’s a complex and technical task, with nuance, and really does require the judgment of an experienced carbon accountant.

Analysing large data sets:

You don’t want to be like me and spend hours perfecting your supply chain analysis via AI, only for it to admit at the end that it had not used actual numbers from my spreadsheet, but estimated what the figures would ‘likely be for a company similar to mine’. After a lot of digging, it admitted that it simply was not equipped to analyze such a large dataset via semantic search (thousands of suppliers). This may work differently with other AIs, but be careful and always be explicit – ‘use actual numbers from the data I provided’, ‘do not estimate’, etc., and build in checks.

Keeping up with regulatory change:

Models have knowledge cutoffs, and are not always up to date on the latest development. The number of times AI would quote the importance of SEC regulations at me… Same with CSRD, even just a week ago AI misquoted the in-scope thresholds to me, stating the old pre-Omnibus figures. Definitely be careful here and do your own research.

Assurance and data integrity:

Assuring AI-derived data may become a problem of the near-future. To avoid any issues, definitely ensure you keep an evidence trail, you can back up (without AI) and calculations performed, and keep record of any spot checking you have done on AI-generated analysis.

Is there a cost-effective AI tool for ESG reporting? (The honest answer)

For now, the general AI tool approved by your company internally is probably your best bet. This means you have to do more work in creating high quality, detailed prompts, and interrogate answers until you know they are right. But in a way this is also good for us – we still have to use our brains and aren’t relying blindly on the technology.

The other tools can be useful in specific cases, and perhaps with time my opinion will change, but nothing beats being able to plug in internal data without worrying about data privacy issues.

How to start using AI in your ESG reporting today

You don’t need a specialist tool or a company-wide AI strategy to get started. If your organization has approved access to a general purpose AI tool like Claude, ChatGPT, or Microsoft Copilot, you can start saving time on real reporting tasks today.

Here are some prompts that will get you started:

For data collection:“I need to collect business travel data from non-specialist colleagues for a 2026 GHG footprint calculation following the GHG Protocol. Draft a short email request and a simple data collection template covering the key data points I’ll need. Make it clear what is necessary vs. what is useful but not absolutely essential, so they’re able to prioritize. My company has 5 offices in different countries, and we use several different travel agencies.”

For supplier screening:“I’m going to give you a list of our top suppliers and a spreadsheet from SBTi showing companies with approved or committed targets. Cross-reference the two lists and produce a table showing which of our suppliers have SBTi status, and calculate what percentage of our total revenue this represents. Please specifically outline in the column whether the supplier has a near-term commitment, near-term target, net zero commitment, or a net zero target. Also please include the years, if target is set. Please note that supplier names may not match exactly, so use your judgment here and if companies appear related (slight variation of name, subsidiary/parent, etc), include them – but make a note so I am able to double check this.”

For disclosure drafting:“Here is our existing climate data, last year’s TCFD disclosure, and the ESRS E1 standard. Draft a first version of our ESRS E1 climate disclosure using this data, flagging any gaps where we don’t have the information required.”

For gap assessment:“Here is our current sustainability report. Cross-reference it against the mandatory disclosure requirements of IFRS S1 and produce a gap analysis table showing what’s covered, what’s partially covered, and what’s missing.”

For consistency checking:“Here are three documents: our sustainability report, our CDP response, and our EcoVadis submission. Identify any inconsistencies in how we’ve described our governance, emissions, targets, or sustainability commitments across the three.”

A few very important notes:

Always check whether your organizations data privacy policy permits uploading internal documents to the tool you’re using. This matters especially for supplier data and financial information.

Always treat AI outputs as a first draft, not a final answer. And be explicit in your prompts — the more specific you are about the standard, the data format, and what you want the output to look like, the better the result. Don’t give it a chance to make any unnecessary assumptions.

We’ll keep updating this as we test more tools and workflows. This space is moving fast enough that what’s true today may not be true in six months’ time. If you’re using AI for sustainability reporting tasks we haven’t covered here, or you’ve found something that works particularly well (or particularly badly), we’d love to hear about it.

Share this post:

Comparison of Terralyn, Claude, and Copilot AI tools tested on sustainability analyst tasks

Terralyn vs Claude vs Copilot: Is a sustainability AI agent worth it?

ESG, road-tested. Terralyn gave us access to their Pro features, so we put it head-to-head against two general AI tools

AI robot waving for carbon footprint calculation.

Can you use AI to calculate your carbon footprint? We tested it!

Depending on your approach to life, you’ve either been (secretly) tempted to use AI for calculating your company’s carbon footprint