Summary: Our webinar explored where AI is delivering results in accounting today, where it still depends on human judgment, and what finance leaders should consider as they govern AI, manage its costs, and reinvest the time it frees up.
Finance teams adopted AI quickly and then reached a plateau. According to Gartner data, the share of finance functions using AI rose from 37% in 2023 to 58% in 2024 and held at 59% in 2025.
Results have been slower to arrive. In a March 2026 survey of 204 finance leaders, 63% said AI implementation in 2025 took longer than expected. The report also highlighted that realized value is not keeping pace with CFO expectations. At the same time, new AI tools for accounting reach the market almost every month, which makes proven capabilities harder to separate from marketing claims.
In the first half of our two-part webinar series, AI and the future of finance: from possibility to practice, we delved into what AI does well in finance today, where it falls short, how to govern it, and what it costs, drawing from our firsthand experience closing thousands of sets of books each month and testing AI tools inside our own workflows.
What can AI do in accounting and finance today?
Transactional accounting
Coding bank and credit card transactions takes time, and every management decision depends on it. Many general ledger platforms now include AI for transaction coding, and add-on tools can provide it for systems without it. Vendors have promoted these features for several years, but we’ve seen them become reliable enough to implement across clients only in the past six months.
Month-end close and reconciliations
AI tools can review the general ledger and compare balances to source documents, such as payroll reports or capitalization tables. Some tools can also draft and post related journal entries. Posting without human review carries risk, so the governance practices described later in this article apply directly to close automation.
Reporting, forecasting, and variance analysis
The largest time savings tend to come from reporting and analysis. Using existing data, AI can produce a working draft of a scenario plan or forecasting model in about 30 minutes, work that previously took far longer to build by hand. Accountants still review and refine the model, but more of their time goes to interpreting what the numbers mean.
Connected systems
Many enterprise resource planning (ERP) systems now include built-in AI for tasks such as transaction coding and journal entry review. Finance teams can also connect their general ledger, bill pay, and customer relationship management (CRM) systems to a general AI assistant and query data across all of them in one place.
Software vendors are also building AI agents into their products. The next stage, in which agents in different systems exchange information and correct each other’s errors, is in development but not yet readily available.
Where does AI fall short in the finance function?
A common AI challenge in finance is an output that is well formatted and confidently presented but reaches the wrong conclusion. Because the error is hard to spot, someone with accounting experience must confirm the result makes sense. We identified three gaps behind the problem:
- Judgment: AI does not apply professional judgment or critical thinking. A reviewer must decide whether the numbers are accurate and whether the conclusion holds.
- Context: AI works only with the data it is connected to and the instructions it receives. It does not know a company’s history, its leaders’ goals, or the relationships behind a decision, unless someone explicitly supplies that information.
- Accountability: AI tools carry disclaimers that leave responsibility for accuracy with the user. A named person must still stand behind every number the finance team releases.
Will AI replace your finance team?
The short answer is no. While AI has changed how finance professionals spend their time, shifting effort from data entry toward review, analysis, and advisory work, decisions ultimately depend on information AI does not have, such as leadership priorities and company history, which keeps experienced people at the center of the process.
How are finance teams using AI in real workflows?
Our webinar makes a distinction between automation, which carries out repetitive, rules-based tasks faster, and AI, which applies logic to work a person would previously have been trained to do. Aprio teams use both, and the following examples show each in practice:
- AI-assisted month-end close: For a fast-growing client, Aprio loaded its standard close workflow and task list into an AI tool. Agents monitor activity throughout the month, such as new fixed assets and changes in equity, and handle recurring tasks such as identifying prepaid expenses and posting payroll entries. The tool then prepares work papers and the monthly reporting package, and a close that once took 15 to 20 days now takes five.
- Lease abstraction for real estate: Commercial leases often exceed 100 pages, and abstracting one by hand took at least three hours. An AI agent trained on each client’s abstract template now extracts the key terms in minutes, and some real estate ERPs can use the output to set up the lease in the system for review.
- Utility bill automation: Real estate companies with properties in many cities and states once relied on staff to log in to each utility’s website and download statements. Bots built with robotic process automation tools now retrieve the bills and route them into the ERP workflow, so team members review entries rather than collect invoices.
In each of these scenarios, the accounting system remains the system of record. AI speeds up how data reaches the general ledger and helps analyze the reports that come out of it.
Each workflow also required setup and testing. Aprio professionals spent more than six months evaluating close automation tools before choosing one, using time studies, accuracy studies, and measures of how much human correction each output needed.
What’s holding back finance teams from adopting AI?
A live poll at the start of the session showed a consistent pattern among the audience. While most attendees said their teams are experimenting with AI or using a few tools, few have applied it systematically across their finance function. Attendees cited not knowing where to start as their biggest barrier. Furthermore, there were pressing concerns about trust, accuracy, and security and compliance.
We explored practical approaches to each:
- Start with low risk, repeatable tasks: Common starting points include drafting emails, summarizing documents, preparing first pass variance commentary, and generating a daily summary of open tasks and unanswered messages. Consistent use on small tasks builds familiarity before teams apply AI to larger processes.
- Strengthen your accounting foundation: Many companies that ask about AI first need documented policies and procedures and a more structured month-end close process. AI produces better results when it builds on consistent processes.
- Plan for change management: Teams adopt AI more readily when they understand why it’s being introduced and how it helps their work. Finance teams already at capacity often need dedicated time or outside support to make the transition.
- Use vetted, enterprise-grade tools: Enterprise versions with appropriate data controls reduce the risk of exposing company or client data. At Aprio, we run every new technology vendor through a formal vetting process for the same reason.
How can finance leaders govern AI and manage risk?
Once AI tools are in use, finance leaders need confidence in data integrity, accuracy, and accountability. Here are a few actionable items to get started:
- Adopt an AI policy: A written policy sets guardrails for which tools the team may use, how those tools connect to company systems, and how leadership expects AI to be applied.
- Keep company data out of model training: Confirm that each vendor will not use your data to train its models and keep company information out of personal AI accounts.
- Manage the agents your team builds: Without oversight, team members can build separate agents for the same task that produce different results. A review and approval process for agents keeps outputs consistent and helps control usage costs.
- Require review and sign-off: AI can be instructed to flag where it is less confident in its own output, which directs reviewers to the items most likely to need correction. A named reviewer should still approve the work before it reaches a client.
Where can finance teams reinvest the time AI frees up?
Asked where they would reinvest saved time, attendees most often chose forecasting, planning, and strategic decision-making. Aprio professionals are moving in the same direction, shifting the main deliverable from the financial statement to the analysis behind it. A variance report, for example, can explain what in the business drove an expense overage and what leaders can do about it, rather than only noting that a line item exceeded budget.
Final thoughts: AI in your finance function
AI already handles transaction coding, reconciliations, and first drafts of financial models in many finance functions. Results depend on a sound accounting foundation, secure tools, a clear policy, and people who review and stand behind every output.
Finance leaders deciding where to start can assess three areas: their team’s comfort with AI, the AI capabilities already built into their current technology stack, and the tasks their team repeats each month. Variance analysis, trend analysis, and a first draft of next year’s budget can be practical starting points, as each uses data the team already has and produces work a person can review before it’s used.
Watch our full webinar replay today for the full discussion.
