As companies rapidly adopt agentic AI for critical financial processes, Sarthak Gupta is urging finance leaders to prioritize internal controls and audit readiness over pure efficiency gains.
Sarthak Gupta, Manager of Transaction Advisory Services at ARDENT Advisory & Accounting LLC, warns that organizations using AI to draft variance commentary, propose journal entries, summarize stakeholder inputs, or develop forecasts for financial reporting are unknowingly integrating the technology into their core financial reporting environment.
“Agentic AI is more than a productivity tool,” Gupta said. “It is now part of the financial reporting process. The key question every organization must answer is: What are the internal controls over and within this system?”
According to Gupta, most enterprises are currently in a “we are figuring that out” phase , a position he considers defensible for the next six months. However, he predicts that by 2027, the lack of robust controls around AI systems could surface as a formal audit finding.
Auditors, he explains, lack a clear classification for AI tools used in financial reporting that fall outside traditional IT general controls or application controls. This ambiguity often leads them to apply the highest-risk lens, placing the burden of defense squarely on the deploying organization.
Critical Questions for CFOs
Gupta recommends that CFOs ask three essential questions before approving further AI investments:
First, what is the audit trail? Gupta stresses that organizations must be able to fully reconstruct AI decisions, identify the data used, and produce documentation suitable for audit scrutiny. “We have the chat logs” is insufficient, he notes, as chat logs do not constitute proper audit evidence.
ALSO READ: BISHEN SINGH URGES INTERNAL AUDIT TO DRIVE INNOVATION AND RESILIENCE
Second, who owns the model? Gupta emphasizes the need for clear ownership by a specific individual within the organization, not the vendor or the AI team, who can be held accountable for unexpected behavior.
Third, what is the validation cadence? Similar to annual revalidation of credit-risk models in banks, AI systems making material decisions require independent, scheduled reviews separate from the deployment team.
“The CFO’s job is not to evaluate AI as a capability. It is to evaluate AI as a controlled, documented, defensible component of the financial reporting environment,” Gupta asserted.
While acknowledging the undeniable productivity gains and competitive pressures driving AI adoption, Gupta cautions that finance leaders who treat AI primarily as a productivity tool rather than a control challenge may face material weakness findings in the near future.
He advises shifting the conversation: capability is what vendors sell, but audit defense is what CFOs ultimately inherit. The finance leaders who will appear most prudent in 2028, according to Gupta, are those who address agentic AI as a control problem first and a productivity solution second.
