How is AI changing accountants' role?
An AI-generated explanation of a financial result can give an accounting team something useful to investigate. Tying that explanation to the business, testing its reasoning, and deciding how to act on it is how accountants shape the value of the technology.
As AI becomes part of the work, those responsibilities evolve alongside it. Barry Melancon, CPA, CGMA, former president and CEO of the American Institute of CPAs (AICPA) and the Chartered Institute of Management Accountants (CIMA), sees data integrity as an important foundation for the profession:1
“Integrity of data is critical, and the profession has a huge role in that. How that looks a few years into the future is going to evolve, and we need to be willing to evolve with it.”
Accountants are in an active position. They help determine where AI contributes, how its output is evaluated, and what people need to learn as their responsibilities change. Melancon emphasizes trust in both the data and what the technology does with it.
For a broader view of the shift, see this overview of how AI is transforming accounting and this look at five accounting trends. For the tension between speed and confidence, the efficiency versus trust takeaways from the 2026 State of AI in Finance report are a useful companion.
How can accounting teams start experimenting with AI?
Give curiosity somewhere to go
Imagine an accounting team considering an approved AI tool to help prepare for a recurring financial review, such as the month-end close. The promise is appealing: less time assembling an explanation and more room to investigate what the numbers mean.
A first attempt produces an observation worth exploring. Perhaps it highlights a change in spending that someone wants to discuss with an operating leader. Following that question helps the team understand the business better.
Experiments like this give curiosity a purpose. The team can examine whether the tool helps people get to a useful question sooner, within the organization's confidentiality, accuracy, and review requirements, including its security expectations. An output that needs correction shows where the data, the instructions, or the review step needs work.
Melancon encourages a willingness to learn while adapting:
“You can’t expect people to be perfect in all of these processes. We have to live in this sort of experimental type of world.”
That willingness still requires care. With a supported trial, people can investigate a new way of working while keeping responsibility for the result clear.
How do accountants make AI-generated work accountable?
Return to the financial review. A polished, thorough explanation provides a helpful starting point. Before AI-generated work informs a recommendation, laying out the information behind it helps everyone understand exactly how it fits the business activity.
Where accountants add expertise
Asking how an answer was produced, examining an unexpected result, and explaining the limits of a conclusion make an AI-enabled process more useful to the next person relying on it. Controls are the practical side of that review. Teams that already use approval workflows and AP controls have a head start on asking who reviewed what before money moves. And this guidance on improving internal controls shows how that discipline is built.
Data integrity sits at the center of that review, because the information behind an AI output shapes how far a team can rely on it. BILL's AI is trained on more than $2 trillion in transactions and more than 1.3 billion documents—anchoring it with a large volume of financial data.2
EisnerAmper's Family Office Services practice offers one example of controls in practice. The practice describes using BILL's audit trails and duplicate bill detection as part of layered fraud controls for the families it serves.3
What reviews can and can't tell a team
Data integrity and controls are central to Melancon's thinking about the profession's role in AI. He also acknowledges that future assurance work brings technical, regulatory, and liability challenges. Routine review and formal assurance have different scopes. That distinction matters—teams must be clear about what it has actually evaluated.
For a firm, that clarity supports a client conversation, and the guide on what changes when AI enters the client relationship goes deeper on that theme. Inside a business, it helps a department leader understand how to use an analysis. In either setting, confidence has something concrete behind it—people who explain the work and take responsibility for its use.
How can AI experiments help develop accountants?
Say an AI tool surfaces an unexpected change in spending. With a manager’s guidance, a newer accountant can check the underlying data, prepare questions, and consider how the answer might affect the analysis. The manager can then review the work, provide their answers, and coach through the reasoning—giving the accountant valuable experience and mentorship.
This way, each AI experiment has two purposes—testing the tool and helping accountants build business judgment and confidence by asking effective questions.
Mentorship opportunities like this will matter more and more as AI changes tasks through which accountants have traditionally learned about business. Melancon points toward mentoring, earlier exposure to senior conversations, and training embedded in real work as ways to build understanding.
Firms working through the people side of this can draw on advice about how accounting firms build their team for growth and creating firm roles that scale. Preparing a question for an operating leader also builds the skill described in how to improve clients' financial analysis.
“I think one of the most valuable skills going forward is the ability to see the big picture, to connect," says Melancon.
Making space for that connection helps a team use new capabilities thoughtfully. Someone learns to investigate a result. A manager learns where coaching is needed. A client or internal partner gains a clearer basis for a decision.
Who does this apply to?
The ideas apply wherever accounting happens: in accounting firms serving many businesses, and in finance and accounting teams inside a single company. The setting differs. The responsibility to evaluate AI's output, explain it, and develop the people involved stays the same.
Putting AI to work: a quick reference
How can a team get started?
- Choose one recurring review where an approved AI tool could help.
- Bring in the people closest to the work.
- Confirm the confidentiality, accuracy, and review requirements the trial must meet.
- Ask a newer accountant to follow one observation through to an operating leader.
- Record what the team evaluated and what it didn't.
- Learn from the results, then adjust and use it to shape the next trial.
Start with one useful question and use it as an opportunity for your team to learn. Experience helps people make better use of AI, as its capabilities and their needs expand. That way, AI keeps learning and evolving alongside your processes and services. Try something small, see what happens, then change with intent.
See AI at work in your financial operations
AI in BILL recognizes patterns in your workflow and captures key invoice fields with 95% accuracy from day one.4 Explore how AI-assisted invoice capture fits into accounts payable automation.
FAQ
How do accountants verify AI-generated financial analysis?
They start by understanding the information behind the analysis and checking whether the explanation fits the business activity.
From there, asking how the answer was produced, examining unexpected results, and explaining the limits of a conclusion make the work more reliable for the next person who uses it. The AP internal controls guide covers preventative policies a finance team can put in place.
Is routine review the same as formal assurance for AI?
No, the two have different scopes. Routine review is how a team checks the work it relies on in daily use, while formal assurance brings its own technical, regulatory, and liability challenges, as Barry Melancon acknowledges.
The practical point is clarity. A team should be able to say what it has actually evaluated.
What should a first AI experiment in an accounting team look like?
It should be small, supported, and tied to one recurring financial review, such as an explanation of a change in spending.
The team works inside the organization's confidentiality, accuracy, and review requirements, and treats any output that needs correction as a signal about the data, the instructions, or the review step.
Does this apply to in-house accounting teams as well as firms?
Yes, the perspective applies to accountants in firms and in companies, since both evaluate AI output, explain it, and develop their people.
This content is presented 'as is,' for informational purposes and is not intended to provide tax, legal or financial advice. Please consult your advisor with any questions.
