one of the most interesting things about AI is not simply what it can do, but what it can make us stop doing. When an artificial intelligence (AI) system produces a polished answer in seconds, the temptation is to rely on it and move quickly to action.
The output may appear authoritative and may draw on more information than any individual could realistically review in the same time. Quite often, it will also be right. This last point matters. The greatest professional risk may not arise when AI produces an obviously poor answer. The harder problem arises when it produces good answers often enough that we stop asking whether this particular answer is reliable, appropriate and suitable for the purpose for which we intend to use it.
The answer may depend on the system’s purpose, the data available to it, its underlying assumptions and how the question or prompt was framed by the user. Professional judgement therefore begins before output is generated and continues through how it is reviewed, challenged and used.
Chartered Accountants Ireland’s guide, The ethical use of AI technologies: a concise guide for members, starts from a straightforward premise. AI can enhance efficiency and insight for Chartered Accountants, but it does not assume responsibility for how its outputs are used.
Accountability for professional work, judgement and decisions remains with the Chartered Accountant.
When AI goes wrong
Public attention naturally focuses on cases where AI produces fictitious information or manifestly unreliable results. Such cases are important, but the underlying failure is familiar. A professional relied on information without adequately checking its source, accuracy or suitability.
The quieter risk is more difficult to detect. If an AI tool performs reliably across hundreds of routine tasks, users may gradually reduce the level of challenge applied to it. The danger is not one dramatic decision to surrender professional judgement. It is the cumulative effect of no longer pausing to exercise it.
The Chartered Accountants Ireland Code of Ethics requires that professional judgement not be compromised by bias, undue influence or undue reliance on others.
A related risk is automation bias: giving disproportionate weight to an automated answer even where professional experience or contradictory evidence provides reason to question it.
The answer is not to reject AI. It is to use it without surrendering the professional disciplines needed to determine whether its outputs can be relied upon.
Where AI adds value
Many valuable applications of AI will never attract headlines. Chartered Accountants are already using it to interrogate datasets, identify anomalies, support technical research, assist reconciliations and prepare forecasts.
These uses can save time and widen the evidence considered, but they also make it increasingly important to understand where AI is being used, for what purpose and how its outputs are reviewed.
The conversation is therefore moving beyond “How can we use AI?” towards equally important questions: Where should we use it? For what purpose? What value will it create? What risks will arise? And what controls and human oversight are appropriate?
These are questions of AI governance.
From experimentation to governance
Good AI governance does not mean putting every use of AI through a committee or turning board members and
finance leaders into data scientists. It means applying familiar governance disciplines to a technology with new capabilities and risks.
We are not reinventing governance for AI. We are ensuring that existing structures for strategy, accountability, risk management, internal control and assurance remain effective when AI enters the process.
Governance should begin with purpose. What is the AI system intended to achieve? Is it suitable for that task? What could happen if it fails, produces an unfair outcome or is relied upon beyond its scope?
Organisations should understand where material AI is being used, assign ownership and assess applications according to their risk and impact. Governance and controls should be proportionate.
An internal productivity tool will not necessarily require the same oversight as an AI system influencing recruitment, credit, corporate reporting, audit evidence or decisions affecting individuals.
Existing policies covering data protection, confidentiality, information security, outsourcing, acceptable use and internal control may need to be adapted.
Oversight should also extend to third-party systems and to “shadow AI”, where employees use unapproved AI tools outside established organisational controls.
The aim should be to integrate AI into the organisation’s governance and control environment, not to create a separate layer of documentation disconnected from how the business operates.
Human oversight must be meaningful
Adding a person at the end of an automated process achieves little if their role is simply to approve whatever the system recommends. Effective oversight requires the time, information, competence, authority and confidence to challenge the output.
It also requires clarity about when a person must intervene, override the system or escalate a concern.
Meaningful oversight connects governance with professional ethics. The organisation must enable effective challenge, while the Chartered Accountant must exercise professional judgement, scepticism and due care.
Trust depends on evidence and challenge
The fundamental principles of the Code of Ethics of Chartered Accountants Ireland are integrity, objectivity, professional competence and due care, confidentiality and professional behaviour. These principles provide a durable framework for AI-enabled work.
As emphasised in the Institute’s guide to the ethical use of AI technologies, AI does not require a separate set of professional values. It creates new circumstances in which familiar principles must be applied.
Consider an AI-generated forecast, and ask yourself:
• Where did the data come from?
• What assumptions underpin this data?
• Is the system being used for its intended purpose?
• Could the data embed bias?
• Does other evidence contradict the result?
• Can the output be explained and defended?
• Am I accepting it because it withstands challenge, or because a sophisticated system produced it?
As AI makes forecasts, reports and recommendations easier to produce, stakeholders will place greater value on knowing how this information has been created, what evidence supports it and who has challenged it.
Chartered Accountants can provide this confidence through disciplined review, transparent explanation and assurance over the information, controls and judgements behind AI-assisted work.
Applying these principles also helps to ensure that efficiency and innovation do not come at the expense of the public interest or the trust placed in the accountancy profession.
The pause that matters
Imagine an AI system has produced an impairment forecast, identified an audit anomaly, proposed a tax interpretation or prepared a recommendation for a board. It has processed thousands of data points and produced an answer almost instantly.
One important step should not be automated: deciding whether this answer should become an action.
This pause is where the Chartered Accountant checks the evidence and assumptions, tests the output for bias or omission, looks for contradictory information, considers who may be affected and accepts responsibility for the conclusion.
It is also the point at which good governance becomes visible. Is ownership clear? Have appropriate controls been applied? Has the decision been documented? Can it be explained, challenged and defended?
AI can increase speed, broaden the evidence considered and reveal patterns that might otherwise be missed. Used responsibly, it can support better professional work.
But the responsibility for the signature, recommendation or decision at the end of the process remains human.
In an economy in which machines increasingly generate answers, the ability to determine which answers deserve to be trusted may become one of the Chartered Accountant’s most valuable skills.
Níall Fitzgerald, FCA/FCPA, is Head of Ethics and Governance at Chartered Accountants Ireland