AI in accounting: Applications, benchmarks, & impact

What is AI in accounting?

AI in accounting is the use of machine learning, natural language processing, generative AI, and robotic process automation to automate or assist with financial tasks, such as data entry, reconciliation, reporting, and analysis.

It doesn't replace human accounting judgment. Instead, it automates much of the manual work that previously came before that judgment.

A report by KPMG Germany published in November 2025 found that 53% of companies are already using AI in accounting or are actively preparing to introduce it, with 61% seeing AI as a decisive success factor in finance.

AI in accounting is no longer a niche experiment. It's the future direction of the profession, even if many firms are still in the early stages of the process.

How AI is used in accounting

AI is most effective in accounting when there are repeatable patterns to learn or documents that need to be processed faster than a person is capable of. The level of risk varies by workflow, and that risk determines how much human oversight is required.

OpenAI's 2025 enterprise report found that accounting and finance users who leverage AI saved more time per message than any other function surveyed, with ChatGPT Enterprise users saving, on average, 40 to 60 minutes per day.

However, those time savings aren't evenly distributed. They're concentrated in structured and repetitive tasks and diminish quickly once a task requires interpretation.

In practice, AI has the power to influence nearly every aspect of accounting, but not to the same extent. The table below summarizes the main accountancy workflows, typical AI use cases, and the required level of human oversight:

Bookkeeping and transaction work

This area is where AI is typically the most reliable because the tasks are high volume and rules-based. AI can take over the manual work that previously took junior staff many hours and produce faster, more accurate records. Common AI tasks include:

• Pulling data from invoices, receipts, and bank statements
• Sorting transactions into the right accounts
• Reconciling bank and ledger balances
• Running payroll and processing bills and payments
• Generating routine financial reports

Audit, tax, and compliance

These tasks need more human oversight than bookkeeping because mistakes can carry legal and regulatory consequences. Using AI can speed up the search-and-review work, but professional human judgment is still needed to make the final decision. Typical use cases include:

• Identifying unusual transactions and potential fraud in large data sets
• Reviewing transactions continuously instead of sampling once a quarter
• Scanning contracts and documents for key terms
• Helping with tax research and preparing standard tax returns
• Tracking changes in regulations

Forecasting and advisory

In this area, AI typically supports human judgment rather than replacing it, and it's where professional accountants add the most value. AI can assist in drafting scenarios, but it's the accountant who must decide what they mean for the client. Common AI activities include:

• Forecasting cash flow
• Creating "what-if" scenarios
• Analyzing results and explaining changes
• Drafting materials for client advisory discussions

Common AI tools for accounting

AI accounting software generally falls into four different groups, each designed to support a different type of work:

• Bookkeeping platforms with built-in AI: Platforms such as Intuit QuickBooks, Xero, and Sage Intacct can manage sorting and reconciliation work within existing accounting systems.
• Spend and workflow tools: Tools such as Ramp can help you automate expense capture, approvals, and bill payments.
• General assistants: Assistants such as Microsoft Copilot and ChatGPT can help with drafting, summarizing, and quick analysis.
• Specialized audit and tax tools: Tools such as Blue J for tax research and the AI features built into audit platforms such as Caseware focus on field-specific review and research tasks.

What are the benefits of AI in accounting?

AI in accounting separates volume from value. By automating repetitive processing, it allows accountants to spend more time on analysis and advisory work.

The benefits of AI in accounting show up consistently in a few key places:

• Accuracy: AI reduces manual-entry errors on high-volume, repetitive tasks.
• Cost: Artificial intelligence reduces the labor hours needed for routine processing.
• Scalability: With AI assistance, a smaller team can handle more clients or a larger transaction volume.
• Speed: AI automates reconciliation and reporting, enabling faster book closures.
• Higher-value work: AI usage allows accountants to spend more time on forecasting, advisory, and judgment-based work.

A Deloitte poll from July 2025 found that around 80% of finance and accounting professionals expect AI-powered tools to become standard in the profession within the next five years. That expectation matters strategically.

Firms that treat AI solely as a back-office efficiency tool risk missing the bigger picture. The real advantage lies in the additional advisory capacity that AI frees up.

What Mercor’s benchmarks reveal about AI's real accounting capability

Benchmarks measure what AI can actually do rather than what vendors claim it can do. One of the clearest public examples is Mercor's APEX AI productivity index, which evaluates AI on real professional work across fields such as finance, consulting, law, and medicine.

The finance benchmark uses investment banking as a reference for complex financial knowledge work because it requires many of the same analytical skills used across the accounting and finance sector. The benchmark asks AI models to complete the kinds of multistep assignments investment banking analysts often perform on the job.

One task, for example, requires updating a merger model to show how different purchase prices and financing choices affect a deal's financial outcome. Performance on these tasks provides a useful indicator of how AI may handle similarly complex accounting work.

APEX results show that today's leading AI models, including GPT-5.3 Codex, GPT-5.4, and Gemini 3.1 Pro, complete about two-thirds of analyst-grade finance tasks on the first attempt, and the latest scores on investment banking analyst tasks show just how quickly AI systems are improving.

These results suggest AI can accelerate complex financial processes. However, experienced professionals are still needed to validate the output and make the final decisions.

You can learn more about how Mercor's APEX AI productivity measures AI models and agents complete real professional work in accounting, investment banking or any other finance related field and why the benchmark is designed to reflect real-world business tasks rather than simple test questions. You can also learn how model outputs are validated before relying on benchmark scores.

How to implement AI in accounting

Successful AI adoption works best as a series of steps, rather than a single rollout. By following these steps, you reduce the risk that AI will introduce errors faster than it will save you time.

Step 1: Identify the accounting tasks AI should handle

Start with high-volume, rules-based work such as data entry, sorting, and reconciliation. Avoid judgment-heavy tasks, such as valuations or complex tax positions, as the evidence shows that AI isn't reliable in these areas yet.

Step 2: Prepare clean, well-structured financial data

Data quality sets the ceiling for the quality of your results. Clean data means having a consistent set of accounts, standardized vendor names, and reconciled opening balances. Messy inputs tend to produce confident yet incorrect outputs.

Step 3: Choose the right AI system to evaluate and keep experts in the loop

Whether you are evaluating an external AI tool, an internal AI agent, or deciding which foundation model to pick for your accounting enterprise workflows, build human review into the process from day one, as research suggests that people tend to over-trust AI-generated answers, even when they are uncertain.

Step 4: Protect client and financial data

Before adopting any AI tool, ask where the data is stored, whether it's used to train the vendor's models, who can access it, and how long it's kept. When working with clients' financial data, the stakes are raised yet further.

Step 5: Test, measure, and scale

Start with a single narrow workflow, and measure performance against a baseline, such as time-to-close, error rate, or hours saved. Then scale only what clearly works. The APEX benchmark family can help you understand how AI model capability varies and APEX Agents benchmark family can help evaluate AI agents' productivity varies across different accounting tasks before you commit.

Risks and challenges of AI in accounting

While there are many benefits of using AI in accounting, each comes with a range of corresponding risks and challenges, which require careful management:

• Hallucination and inaccuracy: AI models can generate plausible but incorrect figures or citations. It's essential to have a human review every output before it's relied on.
• Data privacy and confidentiality: Client financial data can end up in vendor training pipelines without strict contract terms. Always vet data handling processes before onboarding a tool.
• Bias in historical data: Models trained on past categorization patterns can replicate old errors at scale. Regularly audit outputs against known correct samples to ensure accuracy.
• Overreliance: Teams that stop reviewing AI output will lose the ability to identify mistakes. Make human review a standing requirement, not something to phase out.
• Regulatory changes: The Financial Accounting Standard Board regularly issues new accounting standards updates, so it's important to ensure that any AI-assisted work still operates within these evolving regulatory frameworks.

AI governance and controls in accounting and bookkeeping

While risks describe what can go wrong, governance provides the framework to ensure that AI-assisted work remains accurate, compliant, and defensible.

Each control listed below addresses a specific failure it aims to prevent, ensuring the accountant, not the model, remains responsible for the result:

• Human review for AI-generated entries: Check all AI-generated entries and reconciliations before they're posted to ensure errors are caught.
• Audit trails: Log every AI-assisted entry so you can trace it back to who or what created it, ensuring unexplained numbers don't surface during an audit.
• Separation of duties: Keep the person approving AI-generated work separate from the system that produced it. That way, there's no single point of failure between the work and the sign-off.
• Tool and vendor oversight: Test tools before use, track which version you're running, and watch performance over time to catch a drop in quality before it causes harm.
• Clear accountability: Assign responsibility for AI-assisted work to a named professional so that responsibility remains clear.

Future trends in AI in accounting

The trend is clear: AI is becoming more integrated into daily workflows while oversight becomes more formalized.

Expect these AI-driven developments in accounting over the next few years:

• Agent-style tools capable of running multistep processes on their own, such as bill payments or the month-end close
• A "continuous close" where reconciliation happens in real time instead of at the end of the month.
• A bigger emphasis on advisory work, as firms redirect time saved into forecasting and strategy
• Stronger governance as regulators and standard-setters respond to the rapid spread of AI

Will AI take over accounting jobs?

AI is taking over accounting tasks but not replacing accounting jobs. The current benchmark evidence supports that distinction.

AI handles routine, rules-based work well but completes fewer than a quarter of realistic, multiday analyst tasks end to end. That means judgment, client advisory, and sign-off responsibility stay firmly with human professionals.

The trend is shifting toward professionals reviewing and directing AI output rather than producing it manually. New roles emerging around AI are ideal for professionals who understand both accounting and how AI systems actually work.

Getting started with artificial intelligence in accounting

AI in accounting is a powerful assistant for routine, quantifiable work, but it can't replace professional judgment. The firms seeing the greatest gains aren't pursuing full automation. They're automating repetitive tasks, keeping experts responsible for critical decisions, and measuring AI performance on the work that matters most.