KPMG Clara AI shifts audit work toward continuous analysis and judgment

KPMG Clara moves the job from sample testing to broader analysis

KPMG is making a blunt case for how audit work is changing: the firm’s global intelligent platform is built to handle continuous analysis and data-intensive procedures so auditors can spend more time on judgment, skepticism, and complex risk. That matters inside a KPMG audit team because it changes the center of gravity of the job. Instead of spending hours on narrow samples when the data set can support broader review, auditors are being pushed toward pattern recognition, exception handling, and the final call on what evidence actually means.

KPMG describes Clara as a cloud-based smart audit platform that started as a centralized portal, then grew into predictive analytics, and now incorporates cognitive and generative AI. The platform is meant to support risk-based, data-driven execution and deeper coverage beyond traditional sampling. In practical terms, that means the work is less about manually checking isolated items and more about using the system to surface where the real audit risk sits.

The new workflow is built around governed agents, not autonomous decisions

The most important part of the Clara story is not the AI itself, but the guardrails around it. KPMG says Clara’s agents document their reasoning step by step, creating a clear, auditable trail. That is the firm’s answer to the question every audit team has to ask before it trusts a machine: can a reviewer see what the system did, why it did it, and where a human has to step in?

KPMG frames the platform around a Trusted AI framework and a human-in-the-loop mindset. That means the machine is intended to reduce repetitive friction, not to own the audit conclusion. For staff and seniors, the day-to-day shift is real: you are less likely to be reperforming every routine procedure and more likely to be supervising results, tracing the logic behind an output, and deciding whether an anomaly is noise or a genuine risk indicator.

That change also affects review notes and sign-off culture. In a human-led, data-driven audit, a clean checklist is not enough if the explanation is weak. Reviewers need to know why a conclusion is sound, not just that the output matched expectations.

What the rollout says about the scale of the change

KPMG has been pushing this model in stages. In July 2024, the firm said generative AI was being integrated into KPMG Clara for about 90,000 auditors globally. By April 2025, KPMG said AI agents in Clara would empower more than 95,000 auditors globally. That jump is important because it shows the firm is not treating this as a pilot on the edge of the practice. It is moving the workflow model across the audit workforce.

The 2025 rollout included a Financial Report Analyzer AI engine designed to help auditors complete disclosure checklists. KPMG also said additional AI agents would be deployed over the next 12 months into controls testing, financial statement analysis, and related procedures. For a team in the middle of busy season, that means more of the repetitive cross-checking can be automated, but the harder parts of the job do not disappear. Someone still has to decide whether the disclosure is complete, whether the controls story holds together, and whether the financial statements make sense as a whole.

That is where the workload shifts rather than simply shrinks. If the system surfaces more issues faster, the pressure moves to the review layer, where human judgment becomes the bottleneck and the safeguard at the same time.

Regulators still expect skepticism, not blind trust

KPMG’s model fits neatly into a regulatory environment that is increasingly focused on technology without relaxing the core duties of the auditor. The PCAOB says audit risk includes both sampling risk and nonsampling risk, which helps explain why broader, population-level analysis matters. Sampling is only part of the picture, and more data does not automatically eliminate the risk of missing something important.

The SEC approved PCAOB standard updates in August 2024 that address technology-assisted analysis of electronic information, which reinforces the idea that audit work is moving with the tools. At the same time, the IAASB says professional skepticism remains a fundamental element of audits even as technology changes the work. That tension is the real boundary line for Clara and similar systems: technology can widen coverage, but it cannot replace the obligation to question evidence, test assumptions, and challenge what looks too neat.

For auditors inside KPMG, that means the standard for quality is not becoming looser. If anything, it is becoming more explicit. The machine can help identify the outlier; the auditor still has to explain whether the outlier matters.

The skills that matter now are not just technical

This is where the career implications get clearer. In a KPMG environment, especially for anyone thinking about the move from senior to manager or along the partner track, the value is shifting toward people who can translate large data sets into audit judgment. Knowing how to use the platform matters, but so does knowing how to question it, document it, and defend the conclusion in front of another reviewer, a client, or a regulator.

The best auditors in this model will be the ones who can do several things at once: read what the AI flags, understand why it flagged it, distinguish a real risk from a false alarm, and write up a conclusion that a hard-nosed reviewer can follow. That is a different skill mix from the old image of audit as mostly sample testing and tie-outs. It is also more demanding in a subtle way, because the human reviewer is now expected to add value where the machine cannot, not simply confirm what already looks correct.

For KPMG’s audit and advisory professionals, the message is plain. Clara is not making judgment obsolete. It is making judgment more visible, more accountable, and harder to fake.