Finance automation meets reality: AI agents promise speed but CFOs demand accountability

AI Agents Transform Finance Work While Raising Governance Stakes

AI in finance is delivering time savings exceeding 90 per cent in accruals, account reconciliations, and collections work, according to BlackLine1

1

. AI agents now operate around the clock, performing tasks that once consumed finance professionals' hours. Yet this speed comes with a critical requirement: accountability that matches the automation.

Research from ACCA and Chartered Accountants Australia and New Zealand reveals 93 per cent of finance professionals are concerned about the integrity and verifiability of AI-generated insights1

1

. Issues include AI hallucinations, inaccuracies, incomplete data sets, lack of transparency and bias. For CFOs, the question has shifted from whether AI agents can perform financial work to what governance must exist before they trust it to do so.

The Black Box Problem Demands Defensible AI Solutions

Generative AI operates probabilistically, producing outputs based on patterns rather than fixed rules. BlackLine argues this approach needs deterministic controls in AI and financial integrity applications—rules defining what systems can do and providing consistent frameworks for how work gets performed1

1

. An answer that looks plausible fails when results feed into financial processes requiring audit trails.

Defensible AI requires companies to understand not only the work an AI system performs but the instructions it received, what it was permitted to do, and the model used for the task. Those details must be retained so results can be interrogated later. Conversations with regulators, auditors and major audit firms have reinforced the high standard of control and transparency expected around financial information1

1

.

The distinction matters because AI agents in corporate treasury are moving beyond assisting employees to performing longer, more complex financial tasks. A finance professional or auditor should be able to return to a result and establish what information was considered and how the system reached its conclusion—what BlackLine describes as a "glass box" rather than a black box1

1

.

Treasury Departments Face Talent Gap as Institutional Knowledge Retires

A workforce crisis is emerging in treasury departments that has little to do with finding executives. Employees who spent 20 or 30 years processing invoices, reconciling receivables, and managing payments are approaching retirement2

2

. These workers know which systems don't communicate, which exceptions require workarounds, and which processes depend on someone remembering the next step.

Katherine Weislogel, head of Commercial Payments Client Success at Fifth Third, described one client with back office employees averaging 30 years in their positions2

2

. The company wasn't confident it could find people willing to take their places. This institutional knowledge problem is driving practical interest in automating financial workflows.

Commercial payments involve processes that appear straightforward until automation attempts begin. An invoice travels through roughly five separate steps, each potentially involving different people, technologies and decision points2

2

. Experienced employees have served as connective tissue holding these processes together. AI agents offer the possibility of connecting those steps without requiring a person to initiate every handoff.

Finance Automation Shifts From Tasks to Workflows

AI agents can already reconcile accounts, investigate payment exceptions, manage invoices and initiate financial workflows. FIS on October 12 debuted an agentic artificial intelligence solution to help finance teams manage collections, liquidity, risk and cash flow3

3

. The technology moves automation beyond individual tasks to coordinating activity across multiple systems.

One revealing example involves commercial account onboarding. A customer submits documents through a bank portal, but information doesn't necessarily follow the account through the rest of the process. Different departments may request the same information or wait for another team to complete work2

2

. AI agents can keep information with the account as it moves toward implementation, potentially reducing timelines from days to hours.

Yet agentic AI changes the calculation because it introduces discretion. A conventional accounts payable system might match an invoice against a purchase order and route it for approval. An AI agent could investigate a discrepancy, contact a supplier, update records and prepare or initiate payment across multiple systems3

3

. The moment software moves from preparing a transaction to executing one, the economics of automation change.

Permission Management Becomes Critical as AI Gains Authority

Companies are approaching agentic authority incrementally, evaluating permissions based on risks involved. An agent might be allowed to analyze cash positions, prepare forecasts or recommend actions. Giving it permission to initiate or execute a payment raises different questions about fraud, liquidity, entitlements and reconciliation2

2

.

The next employee authorized to spend company money may not be an employee at all. Human employees have identities, job descriptions, reporting relationships and approval limits. Their authority can be documented, modified or revoked. AI agents need the digital equivalent3

3

. Workday on October 13 announced an agreement to acquire C1.ai, a leading agentic identity platform for people and agents.

Paymentology Chief Technology Officer Tim Joslyn noted that "the technology is moving faster than the liability frameworks." The dividing line for automation is reversibility, not complexity3

3

. A defensible record should identify the agent, its originating instructions, the information it relied upon, the permissions in effect and the approvals supporting the transaction.

Real-Time Financial Insights Drive Long-Term Transformation

Finance and accounting teams traditionally spend long hours at month-end gathering information and reconciling accounts before books can be closed. Agentic AI creates the prospect of shifting more work into continuous processing. Instead of concentrating activity around monthly or quarterly deadlines, financial information can be processed as transactions occur1

1

.

The longer-term goal is a continuous close, giving businesses access to real-time financial insights. For Australian companies dealing with exchange-rate movements, tariffs, international trade or intercompany transactions, waiting for the next reporting cycle can leave managers making decisions using information already weeks old1

1

. Substantial sums can be at stake when conditions move between reporting periods.

Yet automation at that scale raises governance questions. A system performing the same task repeatedly across large volumes of transactions can amplify benefits of good decisions but also amplify errors. The PYMNTS Intelligence report "Payment Protection: Why Firms Still Aren't Real-Time Ready" found 70 per cent of surveyed firms plan within 12 months to adopt or expand automation that matches and clears payments against invoices, ledgers and bank statements3

3

.

For mid-market companies, the stakes may be greatest. They have enough transaction volume to benefit materially from finance automation but typically can't afford dedicated security, engineering and compliance infrastructure of multinationals3

3

. The true cost of automation includes not only software and integration expenses but also identity verification, permission management, monitoring, exception handling and financial liability. The next competitive advantage in finance may not be how many AI agents a company deploys but how much authority it can safely delegate through risk management frameworks that prevent the black box problem from undermining trust.