Agentic AI in Internal Audit, What to Automate, What to Keep

 

Consider an illustrative scenario, not a real engagement. A mid-sized distributor connects an audit agent to its ERP, its document repository and its ticketing tool. Within a week the agent has pulled invoices, receipts and approvals for 240 purchase orders and produced a tidy memo saying the procure-to-pay control passed. The audit manager reads it, likes it, and asks the only question that matters: which records did the agent open? Nobody can answer.

The short answer to the sequencing problem is this. Agentic AI should first take bounded, repeatable, evidence-based work: a finite set of documents, stable rules, and output you can check against a known result. It must not decide audit scope, judge whether evidence is sufficient, conclude that a control works, rate a finding, or close an issue. Those decisions belong to accountable auditors, and an agent can only prepare the ground for them.

This article is for internal auditors, chief audit executives and the managers and audit committee members who oversee them. After reading it you will be able to sort audit tasks into automate now, automate with review gates, and keep human. You will also have a short list of access rules, evidence requirements and escalation triggers to put in place before any agent touches a workpaper. The question to ask is not whether AI can replace internal auditors. Ask which audit activities are bounded enough to automate safely, and which decisions need independent professional judgment and a named person who answers for them.

 

SOX Controls Over AI

Consider a composite situation, not a real engagement. A controller opens the monthly bank reconciliation, and the matching tool shows 1,250 unmatched items out of 18,400 bank lines. The accounting manager reviewed every item and approved the package. The external auditor asks which version of the matching logic produced that list, whether the bank file was complete when the tool ran, and what the reviewer saw on screen. Nobody can answer. The matching feature arrived in an ERP update that the supplier switched on, and the SOX risk and control matrix still describes a spreadsheet.

If AI influences financial reporting, your organization must be able to show five things: where the AI is used, how it works, that its inputs and outputs are reliable, that changes to it are controlled, and that a human exercised oversight and left evidence of it. Each of those five becomes an auditor question, and this article takes them in the order a SOX program would address them. Treat the five questions as a practical synthesis of current auditor concerns. No official PCAOB or SEC checklist with five questions exists, and I would be wary of anyone who presents one. The article is written for controllers, SOX control owners, and internal auditors who need working answers before fieldwork starts.

You do not need a separate AI compliance program to get there. COSO released guidance on internal control over generative AI in February 2026 that applies its Internal Control Integrated Framework to generative AI, and it names the risks that matter most to financial reporting: prompt-based manipulation, opaque reasoning, model drift, and frequent configuration changes. SEC staff remarks at the December 2025 AICPA and CIMA conference, as summarized in an EY compendium, pointed the same way, with attention on model design, data, human oversight, and the way AI changes IT general controls. What follows turns that direction into an inventory, a control narrative standard, a validation plan, a change register, and a failure path you can build this quarter.

 

AI Governance Frameworks For Risk Managers

 

Corporate environments are adopting autonomous systems at a pace that outstrips traditional oversight mechanisms. Engineering teams ship automated workflows daily. Business units integrate external models into core operations without formal review. The resulting environment creates massive blind spots for risk professionals. You cannot manage what you cannot see, and you certainly cannot audit what you do not understand. The era of treating artificial intelligence as a mere productivity enhancement is over. It is now a foundational component of enterprise architecture. This shift demands a complete rethinking of control environments.

Risk managers, compliance officers, and cybersecurity experts must transition from passive observers to active architects of machine behavior. The theoretical debates about future capabilities no longer matter. The immediate reality involves managing current deployments that process sensitive data, execute financial transactions, and interact directly with customers. Professionals who master this transition will define the next decade of corporate governance. Those who fail will preside over catastrophic compliance failures and severe reputational damage.

This guide provides a direct, actionable blueprint for securing autonomous systems. We will explore five essential pillars of modern risk management. These pillars move beyond basic policy documents. They focus on the practical implementation of controls, the integration of global standards, and the strategic career positioning of governance leaders. You will learn how to validate machine outputs, secure third-party integrations, assign legal accountability, capture institutional context, and manage the hidden costs of automated code generation.

AI Agent Segregation of Duties Guide for GRC and SOX Compliance

 

Finance leaders are under constant pressure to cut costs, and automation is the fastest lever available. IT teams are asked to hand AI agents more autonomy every quarter, and in many workflows that autonomy now stretches across the full transaction lifecycle: an agent reads customer or financial data, produces a recommendation, approves it, executes the action through an API, and writes the log entry that documents what happened.

No serious organization would hand a single employee that combination of powers without a control wrapped around every step. Yet that is precisely the architecture some companies are building today, one agent deployment at a time, often without anyone deciding to do so on purpose. This piece walks through why that pattern is a genuine segregation of duties problem, what Sarbanes-Oxley actually requires when an agent touches a financially relevant process, and what a control owner can do about it in practice.

The Risk Management Blueprint: A Practitioner's Guide to Quantitative GRC

 

Risk management has a credibility problem. Not because the profession lacks talent, but because color-coded heat maps, ordinal scoring matrices, and quarterly dashboard reviews were never built to change decisions. They exist to document that a compliance process took place. Executive teams know this. They react accordingly by treating risk departments as corporate overhead instead of strategic assets.

I wrote this book to help my peers turn that dynamic around.

After 25 years leading risk functions and advising executive boards across complex multinational companies, I needed a manual that actually bridges advanced quantitative methods with the daily decisions that determine business outcomes. That drive is why The Risk Management Blueprint hit #9 among the most sold risk management books in the weeks after publishing.

At 867 pages, it gives practitioners a single unified methodology across every major risk domain, covering AI systems, cyber exposure, financial cash flows, sustainability transitions, and human behavior. The framework rests on probability theory, financial modeling, and decision science so you can swap subjective scores for numbers that stand up in the boardroom.

You can preview the first four chapters and access the book here: https://amzn.to/4ciag1F

This is not a textbook. It does not spend the majority of its pages diagnosing what is broken in the profession before gesturing toward improvement in a final chapter. More than 70 percent of the book's total length is allocated to domain applications and advanced analytical infrastructure, meaning the bulk of every page is spent on how to build, calibrate, and apply quantitative and predictive risk models across the decisions that actually shape organizational outcomes.

SOC 2 Is Not Security: A Critical Guide for GRC Managers

Let us be direct. SOC 2 is not a security certification. It is an attestation report issued under AICPA standards in which a licensed public accounting firm expresses an opinion on whether a service organization controls met the selected trust services criteria. The report does not declare that your product is safe. It does not certify that your penetration test was rigorous. It does not prove that your attack surface is small or that your customers are protected from a breach. It states that management described a system, selected criteria, asserted controls, and the auditor found those controls suitably designed and, for a Type II report, operating effectively during the specified period.

That distinction matters more than many teams are willing to admit. The report can create a polished artifact that procurement teams accept, but the underlying controls may still be thin, poorly scoped, or disconnected from the technical reality of the product. The phrase SOC 2 is the audit version of trust me bro became popular for a reason. When the PDF is stronger than the security program, the market starts rewarding documentation rather than operational resilience. As a GRC leader, your job is to reverse that sequence. Build the controls, operate them, collect evidence continuously, and let the SOC 2 report fall out as a byproduct rather than serving as the starting point.

This guide walks through the exact gaps that make SOC 2 incomplete as a security signal, how to read a report without wasting your review cycle, how to build a security-first program that makes SOC 2 the output, how to use continuous assurance strategically, and how to govern vendors that hand you a SOC 2 report and expect instant approval. The focus is practical because the risk owner in the room rarely needs another abstract debate. You need a method for using SOC 2 as one piece of evidence among many.

The article is intended for a global audience. SOC 2 is a US-origin AICPA attestation product, but it is now exchanged across borders by cloud providers, software vendors, data processors, and AI product teams. It often sits alongside ISO/IEC 27001, NIST Cybersecurity Framework, CIS Controls, GDPR, HIPAA, NIS2, and emerging AI governance obligations. That means the underlying discipline remains the same. Understand the limitations, own the controls, and never outsource your risk judgment to a report.