Artificial intelligence can compare disclosures, summarise technical material, identify inconsistencies and suggest clearer wording. It can process large volumes of information much faster than a reporting team working manually.
That makes it useful. It does not make it responsible for the annual report.
Corporate reporting still depends on estimates, assumptions, materiality decisions and professional judgement. Someone must decide whether an impairment forecast is reasonable, whether a disclosure is balanced and whether the report gives investors a fair picture of the organisation.
AI can support those decisions. It cannot own them.
This distinction matters for businesses using AI in finance and for candidates preparing for ACCA SBR. AI is not only a technology topic. It is also a reporting, governance, internal control and ethics issue.
Candidates working with an ACCA SBR tutor should be ready to explain both sides of the issue. AI may improve speed and consistency, but management must remain accountable for the information it publishes.
The opportunity is real
Corporate reporting involves a large amount of repetitive and time-consuming work.
Teams compare documents, reconcile figures, check terminology, review previous disclosures and collect information from different departments. They may need to identify changes between multiple versions of the same report or find inconsistencies between the financial statements and the strategic report.
AI can help with many of these tasks.
It may be used to produce an early draft from approved information, summarise a long technical paper, compare current disclosures with the previous year or highlight wording that appears inconsistent across a document.
It may also help reporting teams identify unusual relationships in data. A sudden movement in a balance, a contradiction between narrative and financial information or a missing explanation could be flagged for human review.
Used carefully, this can free finance professionals from some administrative work and give them more time to investigate difficult issues.
That is the strongest case for AI in corporate reporting. It should create more time for judgement, not remove judgement from the process.
A polished answer can still be wrong
One of the biggest risks with generative AI is that an incorrect response may still sound confident and professional.
A weak human draft often looks weak. It may be confused, badly structured or incomplete. That makes it easier for a reviewer to recognise that more work is needed.
An AI-generated paragraph can look finished even when the underlying reasoning is poor.
It may include an unsupported assumption, misunderstand the context or state something as fact when the available evidence is uncertain. It can also produce a general answer that sounds sensible but does not reflect the organisation’s actual circumstances.
This creates a review risk.
A busy finance director may be less likely to challenge a paragraph that already reads well. The quality of the language can create false confidence in the quality of the conclusion.
Reporting teams therefore need to separate presentation from reliability.
A clear paragraph is useful, but it must still be checked against the source information, the applicable reporting requirements and the facts of the business.
AI output should be treated as a draft or analytical prompt. It should not be treated as evidence.
Human judgement begins with knowing what matters
Corporate reporting is not a process of including every available fact.
Management must decide what information is material, which assumptions require explanation and which risks are important to users of the accounts.
These decisions depend on context.
A small operational issue may be immaterial for one organisation but critical for another. A change in a forecast may have limited consequences in a profitable business but create a going concern risk where cash resources are already tight.
AI may help organise information, but it does not automatically understand what matters most to investors in a particular situation.
That judgement must remain with people who understand the organisation, its strategy, its financial position and the needs of users.
For an SBR candidate, this is an important exam point. A good answer should not say only that AI might make an error. It should explain that management remains responsible for deciding what information is material and how uncertainty should be communicated.
Data quality comes before AI quality
A sophisticated system cannot rescue unreliable data.
If the information supplied to an AI tool is incomplete, inconsistent or outdated, the output may reproduce those weaknesses. It may also combine information in a way that makes the original problem harder to see.
Corporate reporting teams often collect data from several systems and departments. Financial information may come from the general ledger, while operational measures, climate information and workforce data may be gathered elsewhere.
Different teams may use different definitions. Information may not have been subject to the same controls as the accounting records. Some figures may be based on estimates rather than confirmed data.
Before relying on AI-generated analysis, management should understand the quality of the information provided to the system.
The reporting team should know where the data came from, who owns it, how it was checked and whether the definitions are consistent.
This is a basic internal control point.
Poor information processed quickly remains poor information. Speed can make the problem more dangerous because it allows unreliable conclusions to spread through the reporting process before anyone challenges them.
AI should support estimates rather than disguise uncertainty
Accounting estimates require judgement because the future is uncertain.
Impairment reviews, provisions, expected credit losses, useful lives, fair values and going concern assessments all depend on assumptions.
AI may help management analyse trends, compare scenarios or process a larger volume of information. It could identify relationships or exceptions that deserve further investigation.
However, it should not make an estimate appear more objective than it really is.
Every model depends on inputs, assumptions and decisions about how information should be interpreted. Those choices remain human choices, even when the final output is produced automatically.
For example, an AI-assisted impairment model may produce a detailed forecast. Management must still decide whether revenue growth is reasonable, whether costs reflect current conditions and whether the discount rate is appropriate.
The board should also consider contradictory evidence.
If the model predicts strong recovery but recent trading remains weak, the optimistic output should not be accepted simply because it was generated through an advanced system.
A good reporting process makes uncertainty visible. It does not use technology to hide it.
The annual report must still sound like the organisation
AI can produce clear and grammatically correct corporate language.
That can be helpful when an early draft is repetitive or difficult to follow. It can also make reports more consistent by identifying different terms used for the same measure.
The risk is that the final report becomes generic.
Investors want to understand how management sees the business. They need specific explanations of what changed, why it happened and what the board is doing next.
A paragraph that could appear in the annual report of almost any company adds little value.
Human judgement is needed to decide where the report should be direct, where uncertainty should be acknowledged and where management should explain an uncomfortable result rather than hide behind vague language.
For example, it is not enough to say that the organisation operated in a challenging environment. The report should explain how those conditions affected revenue, margins, cash flows, forecasts and key estimates.
AI may help improve the wording, but the substance must come from management.
Accountability cannot be delegated
Every significant statement in an annual report should have an identifiable owner.
Someone should be able to explain where the information came from, why it was included and how the conclusion was reached.
The use of AI does not remove this requirement.
A board cannot defend a misleading disclosure by saying that the wording came from a software tool. Management remains responsible for preparing the report, and directors remain responsible for approving it.
This means organisations need clear ownership throughout the process.
The person reviewing an AI-generated section should have enough knowledge and authority to challenge it. A simple confirmation that the text has been read is not enough where the section contains significant judgements.
The reviewer should understand the source information, the assumptions made and the reason for the final conclusion.
Human review must be meaningful. It should not become a box-ticking exercise carried out after the important decisions have already been embedded in the draft.
Explainability matters
A conclusion is difficult to defend when nobody understands how it was reached.
Some AI tools may produce an answer without giving the user a clear view of the reasoning or information that influenced it.
That may be acceptable for a low-risk task such as suggesting alternative wording. It is much harder to accept where the output supports a material estimate or important disclosure.
Management should be able to explain what the system was asked to do, what data it used, what limitations were identified and how the result was reviewed.
The organisation does not need every finance employee to understand the technical design of an AI model.
It does need enough understanding to decide whether the tool is suitable for the task and whether its output can be relied upon.
If a reporting conclusion cannot be explained to the audit committee, the external auditor or a regulator, it should not become more acceptable merely because technology produced it.
Confidentiality creates an immediate ethical risk
Reporting teams handle information that may be highly sensitive.
This can include draft results, forecasts, customer information, employee data, acquisition plans, legal disputes and details of transactions that have not been announced.
Entering that information into an unapproved AI tool could expose the organisation to legal, commercial and reputational harm.
The risk is not limited to deliberately sharing a complete confidential document. An employee may reveal sensitive details while asking a tool to rewrite a paragraph, summarise a contract or explain an accounting issue.
Organisations therefore need clear rules about approved tools and acceptable information.
Staff should know whether prompts are stored, who may gain access to the data and whether the information could be used to improve future versions of the system.
A general instruction to use AI responsibly is not enough. Employees need practical boundaries they can apply during normal work.
Bias and incomplete evidence can distort the answer
AI systems may reflect weaknesses in the data and material used to develop or operate them.
This does not mean every output will be biased. It means the possibility must be considered, particularly where an answer affects people, estimates or disclosures.
A system may place too much weight on historic patterns even when current conditions have changed. It may also produce an answer based on the most common outcome rather than the circumstances of the organisation.
Human review should therefore look for missing and contradictory evidence.
The reviewer should ask whether the output reflects the full range of available information and whether a different reasonable assumption would change the conclusion.
Professional scepticism remains important.
The fact that a result was produced quickly or presented with apparent precision should not reduce the level of challenge.
A practical control framework
Organisations do not need to prohibit every use of AI. They need controls that reflect the risk of the task.
A useful framework should include:
- Approved tools and clearly prohibited uses
- Named ownership for every AI-supported reporting task
- Rules governing confidential and personal information
- Checks over the completeness and accuracy of source data
- Independent human review of significant outputs
- Documentation showing prompts, inputs, changes and approvals
- Escalation where an output cannot be verified or explained
- Training for preparers, reviewers and board members
The level of control should be proportionate.
Using AI to correct grammar in an internal draft does not require the same process as using it to support an impairment assessment or prepare a market-sensitive disclosure.
This distinction allows an organisation to benefit from AI without treating every use as equally risky.
The board needs visibility of actual use
A board may believe that AI is not being used in corporate reporting because no formal system has been purchased.
That does not mean employees are not already using publicly available tools.
Unofficial use can create a serious governance gap. Senior management may have no clear record of which information has been entered into external systems or how AI-generated wording has reached the annual report.
The board and audit committee should therefore ask how AI is currently being used, not only how the organisation plans to use it in the future.
They should understand which activities are permitted, what controls operate and whether management has assessed the reporting and confidentiality risks.
They should also consider whether the organisation’s assurance processes have kept pace with adoption.
Governance should follow actual behaviour. A policy that exists only on paper will not control tools employees already use in practice.
The external auditor still needs reliable evidence
AI may affect both the preparation and audit of corporate reports.
Where management uses AI to support a significant estimate or disclosure, the auditor may need to understand how the tool was used, what information it processed and how management reviewed the result.
A sophisticated output does not become reliable audit evidence automatically.
The auditor should consider the quality of the underlying data, the purpose of the tool, the risk of error and the effectiveness of the controls surrounding its use.
Professional scepticism remains essential.
A polished explanation may still contain an unsupported assumption. An apparently precise analysis may still depend on incomplete data.
The same principle applies where auditors use AI in their own work. Technology may help identify unusual transactions, review documents or analyse large data sets, but the audit team remains responsible for the conclusion.
AI can support audit judgement. It cannot accept audit accountability.
Why this is an ethics issue for accountants
The use of AI connects directly to the ethical duties of professional accountants.
Professional competence matters because accountants need to understand the capabilities and limitations of the tools they use.
Objectivity matters because an AI-generated answer may appear neutral while reflecting biased data or inappropriate assumptions.
Confidentiality matters because sensitive information may be exposed through careless use.
Integrity matters because technology should not be used to produce a more favourable or misleading presentation.
Professional behaviour matters because reporting and legal responsibilities continue to apply regardless of how a draft was produced.
In an SBR ethics requirement, candidates should identify the precise threat and recommend a practical response.
For example, using an unapproved public AI tool to summarise confidential board papers creates a confidentiality risk. Management should stop the practice, investigate whether information has been exposed and require staff to use only approved systems with appropriate controls.
That is stronger than writing a general statement that AI creates ethical concerns.
How SBR candidates should apply the issue
AI is a useful SBR topic because it can connect corporate reporting, internal control, governance, audit and ethics.
However, candidates should avoid producing a general essay about the advantages and disadvantages of technology.
The requirement and scenario should control the answer.
If the scenario says AI was used to prepare an impairment forecast, discuss data quality, assumptions, review, contradictory evidence and management responsibility.
If AI drafted part of the annual report, discuss factual verification, consistency, materiality and whether the narrative reflects the board’s actual view.
If confidential information was entered into an unapproved system, address confidentiality, authorisation, internal controls and remedial action.
If an auditor relied on AI-generated analysis, discuss professional scepticism, audit evidence, explainability and the continuing responsibility of the engagement team.
The answer should then reach a conclusion.
AI may improve the process, but the relevant professional or governing body remains accountable for the final decision.
Write about AI in a board-ready way
A strong SBR answer should sound like advice, not a technology article.
Begin by identifying the reporting or governance issue. Explain why it matters in the scenario. Recommend a control or action. Then conclude clearly.
For example:
Management may use AI to identify inconsistencies in the draft annual report, but the finance team should verify each proposed change against approved source information. A named senior reviewer should approve material disclosures, and confidential information should only be processed through authorised systems.
That paragraph is short, applied and practical.
Candidates following a structured ACCA SBR course should practise this type of scenario-based writing. It is more valuable than memorising a long list of theoretical benefits and risks.
What responsible use looks like
Responsible use does not mean rejecting AI.
It means giving the technology a defined role within a controlled reporting process.
AI can help reporting teams work faster, compare more information and identify issues that require attention. These are genuine benefits.
The organisation must still decide what is material, whether assumptions are reasonable and whether the final report is fair, balanced and understandable.
It must protect confidential information, maintain evidence of review and ensure that important conclusions can be explained.
Most importantly, it must keep accountability with the people responsible for the report.
What to do next
AI will continue to influence how corporate reports are prepared and reviewed.
The organisations that benefit most will not be those that use it everywhere without question. They will be those that understand where it adds value, where it creates risk and where human judgement must remain decisive.
For SBR candidates, the exam lesson is equally clear.
Do not write about AI as if it is only a piece of software. Treat it as a reporting and governance issue. Link it to data quality, estimates, confidentiality, professional scepticism and accountability.
AI can improve the work.
Human judgement must still own the answer.