AI in ESG Reports: How to Integrate AI into an Annual ESG Report
Learn how to include AI in an ESG report, from environmental impacts and supplier data gaps to workforce, customer and governance disclosures. Includes practical examples for Hong Kong and Singapore enterprises.
Lewis Ho

The first question a sustainability team may ask about AI is where it belongs in the annual ESG report.
The answer often emerges somewhere else in the business. HR may be using an AI-supported recruitment tool. Marketing may be using generative AI to prepare campaigns. A facilities team may have introduced software to reduce electricity consumption. Customer service may have deployed a chatbot.
Each team can describe the technology. Fewer may have been asked to document its effects during the reporting year.
That is where ESG reporting begins. An enterprise needs to identify the AI uses connected to its material sustainability matters, establish what it knows about their impacts, and explain its actions and limitations in language readers can understand.
For companies preparing their next annual ESG report, the task need not start with a new standalone chapter. AI can be addressed in existing sections on energy, employees, customers, suppliers and governance. The examples below show how.
Does AI need to be included in an ESG report?
There is no general rule that every ESG report must contain a separate AI section. An enterprise should assess whether its AI uses give rise to sustainability-related impacts, risks or opportunities that are material under the reporting framework applicable to it.
A company using AI to help manage building energy consumption may have information relevant to its environmental reporting. A bank using AI in customer-facing decisions may need to consider fair treatment, privacy and complaints. An employer using AI-assisted recruitment may need to consider workforce and equal-opportunity issues.
The assessment depends on the framework. GRI’s work on digitalization considers impacts on people and the environment, including privacy, bias and resource use. IFRS S1 focuses on material information about sustainability-related risks and opportunities that could reasonably be expected to affect an enterprise’s prospects.
A group reporting in several markets may need to apply more than one lens. It should record which AI-related matters it considered, how it assessed them and why it included or excluded them.

What are the ESG reporting requirements in Hong Kong, Singapore and beyond?
Hong Kong: both Main Board and GEM issuers report
Hong Kong listed companies are required to publish an annual ESG report covering the same period as their annual report. The report may form part of the annual report or be published separately.
HKEX’s Main Board ESG Reporting Code and GEM ESG reporting provisions set out the relevant reporting obligations. Both Main Board and GEM Issuers must disclose Scope 1 and Scope 2 greenhouse gas emissions on a mandatory basis starting from financial years commencing on or after 1 January 2025. For Part D Climate Requirements, Main Board Issuers generally follow a "comply or explain" basis while GEM Issuers are only encouraged to report on a voluntary basis.
For either board, the practical AI question is similar: does an application affect a material environmental, social or governance matter the issuer is already assessing?
Singapore: climate requirements are being phased in
Singapore-listed companies report sustainability information under SGX rules. According to ACRA and SGX RegCo’s updated climate-reporting timetable, all listed companies continue to report Scope 1 and 2 greenhouse-gas emissions from financial years commencing on or after 1 January 2025.
Straits Times Index constituents lead the implementation of other ISSB-based climate disclosures from FY2025, with Scope 3 emissions reporting from FY2026. Other listed companies and large non-listed companies have different timelines.
For a Singapore enterprise preparing its next report, this distinction matters when assessing the environmental effects of cloud and AI services. The company should identify which emissions information it must report, what supplier data is available and which estimates fall within its stated inventory boundary.
Mainland China and other markets
The Shanghai Stock Exchange’s sustainability reporting guidelines apply to specified companies, including constituents of the SSE 180 and STAR 50 indices and companies listed both on the Mainland and overseas. The guidelines consider sustainability impacts as well as financial effects through a double-materiality approach.
Companies with overseas operations may face additional requirements. For example, some Asian groups with European operations may fall within the EU’s sustainability reporting regime, subject to its applicable scope and conditions. The EU has amended that scope, so legal teams should check current rules rather than rely on older implementation summaries.
A group can use a common internal evidence base across markets. It may still need different materiality assessments, presentation and supporting information for different reports.

Seven steps to include AI in your next ESG report
The following steps are a practical workflow for identifying AI-related matters, gathering evidence and preparing disclosures. Not every step will lead to a disclosure. The first six help the enterprise decide what to report and what evidence supports it; the seventh concerns how the ESG report itself is prepared. Together, they help reporting teams connect AI use to existing environmental, social and governance topics.
Step 1: Identify the AI applications that could affect ESG matters
The sustainability team’s first job is to find the relevant applications. The technology department may have an inventory, but business units may also use AI features embedded in software purchased for recruitment, marketing, customer service or facilities management.
Ask each business owner:
What does the application do? Describe the task rather than only the product name.
Who could be affected? Consider employees, applicants, customers and suppliers.
Does it influence a decision? Establish whether it recommends, assists or makes an automated decision.
What information is available? Look for performance measures, complaints, incidents and supplier records.
Which ESG topic could it affect? Link the application to the enterprise’s existing materiality assessment.
A recruitment tool used for thousands of applications may deserve closer examination than a tool used to summarize public documents. The scale of use, potential consequences and available controls all matter.
Sample disclosure: explaining the assessment
Illustrative wording
“During the reporting year, we reviewed selected AI applications against our material ESG topics. The review considered potential effects on employees, customers, data privacy and energy use. Relevant matters were assessed through our existing materiality and risk-management processes.”
Use this wording only if the review took place. If the assessment covered a limited number of applications, the report should say so.
Step 2: Report environmental benefits with a clear calculation
AI may help a company reduce electricity consumption, transport mileage, waste or equipment downtime. Before including a benefit in the ESG report, the reporting team should establish how it was calculated.
Suppose an enterprise introduces AI-supported cooling controls in several offices. Electricity use may fall, but weather, occupancy and changes to operating hours could also affect the result. The company should identify the sites covered, the comparison period and the assumptions used.
Sample disclosure: AI-supported energy management
Illustrative wording
“We introduced AI-supported heating and cooling controls at [number] offices during the reporting year. Electricity consumption at these sites was [amount or percentage] lower than the selected baseline, after adjustments for [relevant factors]. Other operational changes also occurred during the period, and we have not attributed the full change in consumption to the AI system.”
The reporting team should retain the underlying energy data, baseline, calculation method and review records. If it converts the energy change into an emissions estimate, it should document the emission factors and assumptions used.
Step 3: Address the environmental footprint of AI services
The environmental effects of AI are not limited to the benefits an application may deliver. AI services rely on computing infrastructure that consumes electricity and can have other environmental impacts, including water use.
A company operating its own infrastructure may have direct access to some of that information. A company using a third-party cloud service may find it harder to obtain data attributable to a particular AI application.
Sample disclosure: limited supplier data
Illustrative wording
“We do not currently receive sufficient supplier data to estimate the electricity attributable to individual AI applications. We are working with material cloud providers to improve the information available for future reporting.”
This example combines a present limitation with a proposed action. The second sentence should appear only if the company is actually engaging those providers. If it has not yet begun, a narrower statement may be appropriate:
“We do not currently receive sufficient supplier data to estimate the electricity attributable to individual AI applications.”
The enterprise should also explain how relevant supplier impacts are treated within its greenhouse-gas inventory. It should avoid assigning an entire data centre’s electricity use to one application or assuming that a provider’s electricity consumption is automatically the customer’s Scope 2 emissions. The treatment depends on the company’s reporting boundary and methodology.
What Google’s 2025 Environmental Report illustrates
Google’s 2025 Environmental Report provides a reference for discussing the growth and environmental demands of digital services. Google reported that its data-centre electricity consumption increased by 27% in 2024, while data-centre energy emissions fell by 12%. Its total greenhouse-gas emissions increased by 11% from the preceding year.
Those measures have different boundaries. The data-centre figures do not isolate AI, while total emissions cover the wider company. Their value to other enterprises lies in the distinction: electricity consumption, associated emissions and the company-wide inventory can move differently and should be explained accordingly.
Google also reported estimated emissions avoided through certain products. Such estimates describe benefits outside its own emissions inventory. An enterprise making a similar claim should present the estimate separately and explain the comparison and assumptions used.

Step 4: Explain how AI affects employees and applicants
AI-assisted recruitment provides a practical example of how an AI disclosure can sit within an existing social section.
The report should explain what the tool does. Does it help organize applications, recommend candidates or influence a decision? Who reviews its output? Can an applicant raise a concern? What happened during the reporting year?
Sample disclosure: AI-assisted recruitment
Illustrative wording
“We use an AI-supported tool to assist recruiters in reviewing applications for selected roles. Recruiters make the final shortlisting decisions and can override the tool’s recommendations. During the reporting year, we recorded [number] concerns relating to the tool. [Describe any review, correction or other action taken.]”
Before publishing a complaint figure, HR and the reporting team should agree what counts as a concern, how it is recorded and whether the number covers all relevant channels.
Human involvement also deserves a closer look. A statement that a recruiter makes the final decision is more meaningful when the recruiter has the information, time and authority to challenge the recommendation. If the enterprise conducts fairness testing, it should explain what was tested and any important limitations.
Step 5: Describe customer protections and outcomes
A customer-service chatbot can offer faster access to routine information. It can also give an incorrect answer or leave a customer unable to reach a person.
For a material application, a customer section might explain the enquiries handled by the tool, when a case is escalated to an employee and how errors or complaints are addressed.
Sample disclosure: customer service
Illustrative wording
“Our customer-service assistant handles [types of enquiries]. Customers can request assistance from an employee through [available channel]. During the reporting year, we recorded [number] substantiated complaints concerning the assistant and took [describe action] in response.”
If the company reports a change in response or resolution times, it should define the measure and comparison period. Where other service changes occurred at the same time, it should be cautious about attributing the entire result to AI.
This is also a point for the CMO and general counsel to check consistency. A marketing claim about faster service should be supported by the same evidence used in the ESG report.
Step 6: Explain governance through decisions and actions
Readers need to understand who is responsible for material AI uses and what happens when a concern arises. A governance disclosure can cover management accountability, review before deployment, supplier assessment, escalation and remediation.
Sample disclosure: review before deployment
Illustrative wording
“AI applications that may significantly affect employees or customers are subject to documented review before deployment. During the reporting year, we reviewed [number] proposed applications. [Where accurate: One application was paused while the business owner addressed issues identified during testing.]”
The example becomes useful when the enterprise can show what the review covered and who made the decision. If no application was paused, the report should not imply otherwise.
The voluntary NIST AI Risk Management Framework may help enterprises structure their AI risk processes. ISO/IEC 42001 sets requirements for an AI management system. Neither replaces the enterprise’s applicable ESG reporting requirements.
A banking example: DBS
DBS provides a non-technology-sector reference. In its 2025 Sustainability Report, the bank discusses its use of AI to enhance climate and ESG risk assessment. In a separate discussion of responsible AI in banking, DBS describes four principles in its model data governance framework—purposeful, unsurprising, respectful and explainable—and says a senior-level committee oversees AI use cases.
The two sources serve different purposes. The sustainability report connects AI to an existing ESG risk process; the separate governance discussion explains the bank’s approach to AI use. Together, they show how an enterprise can give readers a route from the sustainability issue to the relevant management framework.
Another financial-services example: Interbank
In its FY2025 ESG supporting document, Intercorp Financial Services describes Interbank’s approach to responsible AI. It discusses assessment and monitoring practices intended to identify potential bias and says the bank plans to strengthen its measurement framework over time.
The document also describes a responsible-use manual for Claude, including restrictions on information handling and security guidance. That is a useful reminder for enterprises preparing their own reports: governance may involve general principles, but it also depends on instructions for the tools employees actually use.
Step 7: Put controls around AI used to prepare the ESG report
An enterprise may use AI to search approved documents, summarize supplier responses or draft sections of its report. Those uses can save time, but generated text still needs verification.
A reporting process should require a named reviewer to check material statements, figures and calculations against approved sources. The team should also establish which tools may receive confidential, commercially sensitive or personal information.
Sample disclosure: AI in report preparation
Illustrative wording
“We used approved AI tools to assist with selected research and drafting tasks for this report. Material disclosures and calculations were reviewed by the relevant business owners against supporting records before publication.”
This wording is suitable only where the review was performed and documented. A company should not claim that every output was checked unless its process supports that statement.
A practical checklist for the next reporting cycle
The work can begin before the ESG report is drafted:
Identify the applications. Ask technology, HR, customer service, marketing, procurement and facilities teams where AI is used.
Assess materiality. Consider the people affected, environmental consequences, scale of use and relevance to existing ESG topics. Record the decision and the framework used.
Assign an owner. Each proposed disclosure needs someone who can explain the application and produce supporting records.
Define the measure. Establish what is being counted, the period covered, the baseline and any limitations. A complaint, a confirmed error and a privacy breach may require different definitions.
Request supplier information. Ask material providers what data they can supply about the services used. Record gaps rather than filling them with unsupported estimates.
Review the draft across functions. Sustainability, legal, risk, finance and communications should check the evidence, wording and consistency with other public claims.
Keep the record. Retain calculations, source documents, decisions and approvals in a form that can support internal review or assurance where applicable.
Preparing an AI disclosure your enterprise can support
A useful AI disclosure may be a paragraph in the workforce section, a note explaining a cloud-data gap, or an energy case study with a clear baseline. Each should be supported by records showing what the enterprise used, what it observed and what action it took.
LexGuard AI helps enterprises build the governance records behind those disclosures. We work with legal, risk and business teams on AI policies, vendor due diligence, contract protections, data governance and executive training. These controls can help organisations identify AI-related risks, document decisions and strengthen the evidence available for ESG reporting.
Contact LexGuard AI to discuss your organisation’s AI governance and reporting readiness.

Does an ESG report need a separate section on AI?
Not necessarily. An enterprise should assess whether its AI applications affect material ESG matters under the reporting framework applicable to it. Relevant information can often be included in existing sections on energy, employees, customers or governance. Corporations should check the requirements applicable to their board and reporting year.
How can a company include AI in its ESG report?
Start by identifying AI applications that could affect material ESG matters. Then gather evidence of their impacts, controls and outcomes. For example, a company might report on an AI-supported energy project, oversight of an AI recruitment tool, or safeguards for a customer-service chatbot. Claims should be supported by records, and disclosures should explain relevant limitations.
How should a company report AI’s environmental impact when supplier data is unavailable?
Explain the data gap rather than publishing an unsupported estimate. For example, a company could state:
“We do not currently receive sufficient supplier data to estimate the electricity attributable to individual AI applications.”
If it is engaging cloud providers to improve the information available, it can also describe that work. The company should explain how relevant supplier impacts are treated within its greenhouse-gas inventory, using its applicable reporting boundaries and methodology.
