Artificial intelligence is becoming increasingly involved in corporate accounting and finance operations, with capabilities extending beyond research, reporting, and financial analysis. As organizations begin exploring AI agents that can complete tasks, investigate discrepancies, and execute portions of financial workflows, CFOs and controllers must determine where greater automation makes sense and how to maintain appropriate financial oversight.
During a recent Controllers Council and Paystand webinar, Beyond the Copilot: When AI Starts Doing the Work of Finance, Christine Gu, CFO at Clarify Health Solutions, and Philip Peck, SVP Finance Transformation at Peloton Consulting Group, discussed the practical applications of agentic AI, the risks associated with granting AI greater autonomy, and the responsibilities finance leaders must consider as these technologies become more prevalent.
The discussion examined opportunities across accounts payable, accounts receivable, financial planning, reconciliations, and reporting, along with the importance of internal controls, data integrity, and professional judgment.
Where Finance Organizations Stand with AI Adoption
Although AI adoption continues to expand, many finance organizations remain focused on applications that assist employees rather than independently execute financial tasks. However, interest is growing in technologies that can participate more directly in established financial processes.
Gu described three levels of AI adoption within finance organizations. The first involves copilot-style assistance, including drafting memos, summarizing contracts, and explaining financial variances. The second incorporates AI capabilities directly into existing applications, such as automated invoice coding, anomaly detection, and recommended follow-up activities.
The third level involves agentic AI, where systems can execute multiple steps within a financial workflow under defined permissions.
For organizations considering this progression, Gu emphasized the importance of examining existing processes before introducing additional automation. Finance leaders should understand which activities require substantial manual effort, which workflows are repeatable, and where human approval remains necessary.
Peck added that organizations are increasingly interested in incorporating AI directly into enterprise systems and established financial processes. The objective is to improve operating efficiency, shorten processing cycles, strengthen controls, and support more informed financial decisions.
Rather than introducing AI across numerous functions simultaneously, finance leaders should evaluate where the technology can address specific operational difficulties and deliver measurable improvements.
Identifying the Strongest Opportunities for Agentic AI
Accounts receivable, accounts payable, month-end close, and financial planning are among the functions where agentic AI could have a meaningful operational impact.
Gu identified accounts receivable, collections, accounts payable, and selected month-end close activities as particularly suitable starting points because they involve substantial transaction volumes, repeatable procedures, and measurable outcomes.
In accounts receivable, for example, an AI agent could examine aging reports, identify overdue accounts, prepare customer communications, investigate payment disputes, and direct exceptions to the appropriate employees. Finance professionals would continue managing customer relationships and decisions requiring additional discretion.
Within accounts payable, AI agents could extract invoice information, identify duplicate invoices, match transactions against purchase orders, and assemble payment proposals. However, Gu emphasized that sensitive activities, including releasing payments and modifying vendor banking information, should remain subject to independent human approval.
Similar opportunities exist within the financial close. AI could collect supporting schedules, investigate reconciliation differences, and prepare documentation for review.
Peck agreed that high-volume processes with established procedures provide a practical foundation for agentic AI. He also identified financial planning and analysis as an area with considerable long-term potential.
AI agents could eventually support more continuous forecasting by monitoring revenue, expenses, headcount, project spending, and cash flow between traditional planning cycles. These systems could investigate changes, examine underlying business drivers, evaluate alternative assumptions, and recommend adjustments.
Such capabilities could give finance teams more time to interpret financial performance and advise business leaders, provided that significant assumptions and consequential decisions remain subject to professional review.
Why Internal Controls Must Remain a Priority
Allowing AI to execute financial transactions introduces additional considerations involving accuracy, security, authorization, and accountability.
Peck explained that financial organizations must consider the consequences of incorrect or unauthorized actions. An AI agent might misinterpret an invoice, select an incorrect accounting period, initiate a duplicate payment, or act outside its designated authority.
More complicated risks arise when agents execute several connected activities. For example, allowing one agent to create a vendor, process an invoice, and initiate payment could compromise segregation-of-duties requirements.
Errors could also spread across interconnected workflows, particularly when multiple AI agents rely on the same inaccurate information. Without appropriate monitoring, a single incorrect assumption could affect numerous transactions or subsequent financial activities.
Gu identified accountability and auditability as central considerations when evaluating AI applications. Before allowing an agent to operate within a production environment, she recommended addressing four questions:
- What information can the agent access?
- What actions can it perform without approval?
- How will errors be detected and resolved?
- Can the organization reconstruct the agent’s actions for an auditor?
These questions help establish appropriate permissions, approval requirements, monitoring procedures, and supporting documentation.
Gu also emphasized that the underlying control framework should reflect the financial action being performed, regardless of whether a person or an AI system completes the activity.
Organizations should therefore incorporate least-privilege access, independent verification, exception monitoring, and clear escalation procedures into their AI implementations.
Who Should Lead AI Adoption in Finance?
The panelists agreed that successful AI implementation requires coordinated leadership across finance, technology, security, and operational teams.
Gu identified the CFO as the appropriate executive sponsor responsible for establishing strategic priorities and determining the business case for AI investments.
However, implementation responsibilities should extend across several functions. Controllers must establish financial controls and documentation requirements, process owners should remain accountable for operational outcomes, and IT and security professionals must evaluate technical architecture, system access, and data protection.
Gu recommended establishing a small governance committee to coordinate these responsibilities rather than assigning complete ownership to a single AI specialist.
Peck supported this approach and suggested that organizations also involve internal audit, compliance, and data management professionals where appropriate.
External advisors may assist with evaluating existing capabilities, establishing implementation priorities, developing deployment plans, and supporting change management. The extent of outside involvement will depend on organizational size, complexity, and available expertise.
Both panelists also emphasized the importance of improving AI literacy throughout the finance department. Employees must understand how to evaluate AI-generated information, question potentially inaccurate results, and apply established accounting principles when reviewing automated work.
Building a Business Case for AI in Finance
When discussing how CFOs and controllers can achieve meaningful results from AI investments, both panelists recommended beginning with existing business problems.
Peck encouraged organizations to approach AI adoption as part of broader finance transformation efforts, with particular attention to process design, internal controls, and measurable financial outcomes.
Gu recommended selecting one or two repetitive processes that consume substantial employee time and establishing performance benchmarks before introducing AI.
Potential measures include processing time, manual intervention, error rates, and operating costs. Organizations can then conduct controlled pilots and compare the results against their original benchmarks.
Gu also stressed the importance of correcting underlying process deficiencies before introducing automation. Inconsistent master data, unclear approval responsibilities, and poorly coordinated handoffs can undermine AI implementations regardless of the technology’s capabilities.
Another consideration involves how organizations account for AI investments in workforce planning.
As AI assumes additional operational responsibilities, finance leaders may need to evaluate technology expenses alongside traditional staffing costs. This approach allows organizations to compare hiring, process redesign, and automation when determining how to meet future business requirements.
Gu suggested that AI costs should become part of workforce budgeting as organizations prepare their financial plans. Evaluating these expenses together can provide a more complete understanding of future operating costs and staffing requirements.
How AI Could Change Accounting Careers
The webinar also addressed how AI may affect accounting responsibilities, professional development, and future talent requirements.
As transaction processing, reconciliation preparation, and report compilation become increasingly automated, accounting professionals may spend more time monitoring exceptions, validating financial information, reviewing controls, and interpreting results.
Gu cautioned against if greater automation will necessarily eliminate accounting positions. While routine activities may require less manual involvement, responsibilities involving complex judgment, communication, internal controls, and business advisory work will continue to require experienced finance professionals.
However, Gu raised an important concern about developing entry-level accounting talent.
Junior accountants traditionally gain experience by preparing reconciliations, tracing transactions, and completing foundational accounting assignments. If AI assumes much of this work, organizations will need alternative methods for teaching employees how financial systems operate and how to exercise sound professional judgment.
Gu and Peck both emphasized the importance of developing training approaches that preserve accounting knowledge and prepare future controllers and CFOs.
The discussion also addressed emerging governance responsibilities. As AI becomes more involved in financial operations, finance leaders may assume additional duties involving AI governance, data management, cybersecurity, and automated process oversight.
These responsibilities will require continued cooperation between finance, technology, and risk management teams, along with ongoing attention to the policies and procedures governing financial information.
Preparing Finance for Greater AI Autonomy
Agentic AI presents opportunities to reduce manual processing, improve financial analysis, and support more responsive decision-making. However, the panelists emphasized that successful adoption depends on selecting appropriate applications, establishing dependable controls, and maintaining clear accountability.
For CFOs and controllers, the immediate priority is to evaluate existing workflows, identify activities suitable for automation, and determine which decisions require continued human involvement.
Organizations that establish reliable data practices, measurable performance objectives, and well-defined approval procedures will be better positioned to expand AI capabilities while preserving the integrity of their financial operations.
As finance responsibilities evolve, professional judgment, accounting expertise, and effective governance will remain essential to ensuring that AI contributes to reliable financial outcomes.
Watch the Full Webinar
To hear the complete discussion with Christine Gu and Philip Peck, including practical examples of agentic AI in finance, internal control considerations, and recommendations for CFOs and controllers, watch the full webinar.
ABOUT OUR SPONSOR:
Paystand is the modern, digital B2B payment network that seamlessly integrates with agentic accounting and finance. Paystand’s mission is to be a one stop shop for Controllers and CFOs with AR, AP, Payments, Spend and Expense management, and much more. Learn more at www.paystand.com
