Artificial intelligence (AI) agents have become a mainstream aspect of finance and accounting. While agents are a productivity revelation, they do not change accounting fundamentals. Even if an agent performs a task, a person has to be responsible for it and the more judgement required, the closer a human needs to be to the task. This beings with appropriate controls and oversight.
With AI agents, the right level of human involvement is a balance. Too much involvement risks giving up the time and efficiency gains agents offer professionals. Too little and the risk is mistakes. The crux is making sure speedier workflows don’t compromise quality.
The finance departments that use AI effectively will have a significant advantage. It’s is just important not to assume that once installed, an AI agent will continue to produce correct results. Their output needs to be continually checked.
Agents take action
The difference between artificial intelligence and AI agents is agents can act. AI tools like ChatGPT and Claude answer questions, generate content and solve problems like planning itineraries or producing background research. Ask a question and receive a high-quality response in seconds. While AI tells users what to do, AI agents perform the task. In finance, this must be subject to permissions and controls.
“Responsible use starts with understanding the level of risk associated with the task being performed and designing appropriate controls around that task. The more complex the task, the stronger the controls need to be,” says Nick Perrett, CEO, Yarra Lane Group.
Perrett has spent four years designing and implemented agentic AI systems and is currently exploring how human judgement, robotic execution and AI agents can work together.
As an example of how controls can be implemented, if an AI agent is used to prepare tax research, the agent can be restricted to drawing on approved sources, required to reference specific legislation and told to provide links to any supporting material it has used to arrive at any conclusions. The agent can also be instructed to identify the professional standards it has considered when producing any advice and to document any assumptions, risks and alternative views it has considered. The system can then prepare the output in a Word document, together with supporting workpapers for review.
While the major accounting platforms are rolling out agents now, it’s possible to build AI agents through platforms like ChatGPT. New tools are also emerging.
Praxio AI has developed an AI app which helps with tax research “When we built our practical tax assistant, we only used legislation and ATO guidance or ATO documents and nothing else,” says Praxio AI adviser and content creator, William Young.
With all the enthusiasm around AI, the risk is staff will use them without approval or the business’s knowledge. Education is key here.
“Staff need to understand the opportunities and risks. From a practical perspective, I believe finance teams should provide approved, enterprise-grade AI tools rather than simply prohibiting their use,” says Perrett.
“If you ignore AI, employees will seek out their own solutions using free consumer products, which can create a much greater security and compliance risk. The objective should not be to prevent the use of AI, but to provide secure tools, clear policies and appropriate training so it can be used responsibly,” he says.
A question of constraint
AI agents are not just a smarter spreadsheet, they are delegated actors making decision and executing and they need appropriate oversight. Most important, the way agents are overseen has to be figured out before they are adopted. Ideally, a strong AI governance framework supports a system through which every significant action an agent takes, source it draws on and assumption and decision it makes should be capable of being reviewed.
“The control question has changed. It is not enough to ask whether a human reviewed the output. The better question is whether the agent was authorised to perform the exact action, within exact limits, using approved data, code and policy and with a named human accountable,” says David Lee Kuo Chuen, professor, school of business, Singapore University of Social Sciences.
The objective is to design a governance system so any failure is immediately surfaced by the system and actioned. This means constantly monitoring agents’ outputs.
“That is why I would like to move from the phrase human-in-the-loop to a clearer delegation-of-authority model. Every finance agent capable of material actions should have an identity, human sponsor, permitted purpose, a financial authority limit, approved tools, an expiry date and a revocation mechanism. The firm should know who the agent is, what it can do, what it cannot do and who is accountable when something goes wrong,” Lee says.
A clear delineation of tasks between people and agents is required.
“An AI agent can draft a memo, prepare a first-pass analysis or recommend a journal entry. But a named human should approve material, irreversible or external-facing actions,” says Lee. Examples include regulatory filings, impairment judgements, revenue recognition judgements, payment releases or material journals.
Ordinary system access controls may not be enough if an agent can use a browser, click buttons and move through systems like a person. “The agent should operate in a sandbox, with domain allow-lists, session logs, action-by-action approval where needed and alerts when it steps outside its mandate,” says Lee.
National regulatory postures around AI are emerging. Australia’s Guidance for AI Adoption sets out the AI6, six essential practices for responsible AI governance and adoption by organisations operating in Australia. The checklist covers accountability, understanding impacts, managing risks, sharing essential information, testing and monitoring and maintaining human control.
In financial services, the Australian Prudential Regulation Authority’s (APRA) AI expectations make it clear AI is an operational resilience, cyber and board-oversight issue for the entities it regulates. APRA is among many regulators looking at AI and the Australian Securities and Investments Commission has raised concerns about the use of AI in businesses outpacing updated governance frameworks.
Singapore’s Model AI Governance Framework is instructive as it was written with agents in mind. The framework recommends placing limits on agents’ autonomy and introducing checkpoints where human approval is required.
“The requirement to assess and bound risks upfront means classifying AI use cases by materiality, autonomy, reversibility and data sensitivity. Making humans meaningfully accountable requires assigning a CFO sponsor, controller, agent owner, data owner and internal audit responsibility,” Lee says.
Above all, the legal obligations professionals need to meet remain the same, whether or not an agent is involved in a finance team’s work. “Boards, CFOs, auditors and lawyers still need to decide who has authority, what counts as approval and how liability is allocated,” Lee says.
Practical matters
Even when AI agents are tasked with a problem, professional standards require accountants to understand and document how any conclusion is reached.
“That principle should not change because AI is involved. If an existing process needs a manager to review work before it is provided to a client, the same review process should apply to work produced by an agentic AI system. In many cases, the supporting documentation generated by AI can be more comprehensive than the workpapers traditionally produced by staff because references, sources and reasoning can be captured automatically,” says Perrett.
This means systems need to be implemented to verify AI outputs in the same way as work produced by any other team member.
“If a task currently needs to be reviewed and signed-off, that will be the case even if it’s performed by an agent. The reviewer should assess the quality of the work, validate the conclusions, examine the supporting evidence and determine whether the output is appropriate for the circumstances,” says Perrett.
Using AI agents responsibly is not set-and-forget exercise.
“You need to have policies and controls in place, then you should be performing continuous testing in the organisation to make sure people are using it properly and risks are being managed,” says Young.
For finance teams that are experimenting with AI agents, one approach is to treat a new agent like a new colleague.
“Qualify it, understand how it does the job and test it on work where you already know the right answer, before you let it touch anything live. Check it closely in the early months, the way you would review a new hire under probation before trusting them unsupervised. Keep a periodic control check running on the system, the same as any internal audit. To effectively implement AI agents, certain parts of your processes have to change,” says Bryan Sng, co-founder and COO of AI agent accounting platform, SimpleAI.
“Verify by reconciling the output back to an independent source, the bank statement or the source invoice, not by judging whether it looks right on screen. Ease off oversight only when you have evidence the agent is reliable, not because three months have passed. Keep a regular check on the agent’s output to ensure that it is working as desired,” says Sng.
Spot checks and full, regular reconciliations are also essential.
“Randomly select accounting vouchers, journal entries and report items to compare AI outputs with original source data and reconcile total accounts to check data consistency,” says Collin Jin, Deloitte China audit and assurance innovation and digital services leader and president of CPA Australia’s east and central China committee.
“Then, apply rule-based, cross-verification. Embed accounting standards, accounting policies and logical formulas into inspection rules to spot abnormal entries, mismatched figures and unreasonable analytical conclusions automatically,” says Jin.
Two staff should be responsible for checking the output. Allow junior staff to conduct primary verification, while senior financial professionals can do a secondary audit, especially with financial statements and critical analytical results. To make sure the system is working. track model performance continuously and record error rates regularly, then retrain and optimise AI models when deviations happen.
It’s good practice to carry out scenario and simulation tests. “Use historical and simulated business data to verify AI adaptability under complex conditions. Such multi-layered checks ensure the accuracy, compliance and reliability of AI-generated financial outputs,” says Jin.
The bigger picture
Any fully autonomous AI or agent system should be approached with caution.
“While AI can deliver impressive speed and accuracy, finance professionals operate in an environment where privacy, confidentiality, regulatory obligations and professional judgement are critical. AI systems can and do make mistakes. More importantly, some of the decisions we make involve subjective judgement where there is no single correct answer,” says Perrett.
Figuring out what should be done by an agent, human and robot and how they work together is where the work is now. In the long term, every finance team will have access to similar AI tools. The differentiator will be how they use that technology.
Read all about in the September 2026 edition of CPA Australia’s In the Black magazine.
https://intheblack-magazine.cpaaustralia.com.au/intheblack-september-2026?pid=ODk8959222&p=43&v=1.1