KPMG AI in Auditing: Pilot of Next-Generation Audit Agents Begins
Audit fees were lowered by 14% after the firm highlighted AI efficiency benefits.
KPMG launched an AI-enabled audit and advisory offering for private companies last week, the firm’s latest move in a multi-billion-dollar push to embed artificial intelligence across every part of its business. The new product, called KPMG Private, pairs the firm’s Clara audit platform with integrated tax and advisory services aimed at a private market that KPMG AI leaders say is growing more complex and facing expectations once limited to publicly traded companies. Global private market assets are projected to exceed $18 trillion by 2027.
“KPMG Private brings the full strength of our firm to our clients, seamlessly connecting our sector expertise, global capabilities, and leading technology,” Tim Walsh, KPMG LLP’s chair and CEO, said in a statement. The offering includes agentic AI tools for data collection and reporting, a dedicated private company audit practice running on Clara, and access to KPMG’s Digital Gateway for tax services. According to the firm’s own data, 81% of private companies are already piloting or using AI in accounting and financial planning.
The launch extends a buildup that has accelerated sharply over the past 18 months. KPMG first embedded AI agents into Clara last spring, targeting expense vouching, identification of unrecorded liabilities and assessment of accrued expenses. It also released a Financial Report Analyzer engine that helps auditors complete required disclosure checklists. At the time, the firm committed to deploying a second wave of agents focused on controls testing and financial statement analysis within 12 months. That deadline falls this summer and will represent the deepest integration of KPMG AI into the core of the audit process.
Results from the first round of KPMG AI agents are already in. Pilot engagements across banking, consumer goods and energy clients showed a 35% reduction in hours spent on vouching and workpaper assembly, according to KPMG’s internal tracking data. Audit teams using the agents analyzed 100% of revenue and expense line items, compared with the 5 to 10% sample that is standard practice. The expanded coverage surfaced duplicate-invoice schemes that had gone undetected in prior cycles.
Thomas Mackenzie, KPMG’s U.S. and global audit chief technology officer, said Clara’s agents now identify which expenses require testing, ingest the source documentation electronically, extract the data, generate the workpaper and run the testing automatically. The auditor reviews the output and applies judgment rather than performing the comparison manually.
The broader infrastructure is in place. KPMG launched Workbench in mid-2025, a multi-agent platform running on Microsoft Azure AI Foundry Services with 50 AI assistants in operation and nearly a thousand in development. Workbench sits beneath Clara, Digital Gateway and KPMG Velocity, connecting audit, tax and advisory on a single foundation. A separate deal with Google Cloud will use the Agentspace platform and the Agent2Agent interoperability protocol to serve joint clients, giving KPMG AI access to multiple model providers. The firm became the first organization globally to earn BSI/ISO 42001 certification for AI management systems, and every agent on its platform is assessed against its 10-pillar Trusted AI Framework.
All four of the Big Four are now in the same race. Deloitte rolled out Zora AI with Nvidia and expanded generative AI in its Omnia audit platform. EY gave 80,000 tax professionals access to 150 specialized agents, with plans to reach 100,000 by 2028 and more than $1 billion in annual AI spending. PwC introduced Agent OS and has deployed 25,000 agents across client operations. The technology is not without risk. Deloitte Australia agreed last October to partially refund the government after errors were found in a report produced in part using AI.
The economics of KPMG AI are also coming under scrutiny. KPMG International negotiated a 14% fee reduction from its own auditor, Grant Thornton UK, arguing that AI should make the work cheaper, the Financial Times reported earlier this year. Grant Thornton agreed to the cut, according to filings at UK Companies House. The move gave every audit client in the market a reference point for renegotiating fees. EY’s global managing partner for growth and innovation, Raj Sharma, has separately suggested that AI agents could push the profession away from hourly billing toward a “service-as-a-software” model based on outcomes.
Regulators have yet to provide a clear framework. Christina Ho, then a PCAOB board member, warned in a speech last fall that if auditors use AI to test 100% of journal entries instead of sampling, inspectors could either credit the broader coverage or require such detailed documentation of the AI model that firms revert to sampling to reduce compliance risk. Ho, who has since left the board, called for the PCAOB to develop risk management guidance and create an Innovation Lab to work with firms and technologists. The board has listed generative AI use in audits as a formal inspection priority, a designation that puts the KPMG AI rollout directly in regulators’ line of sight.
The profession’s talent shortage adds urgency. More than 300,000 accountants and auditors have left the field since 2020, according to the Bureau of Labor Statistics, and the AICPA reports that 75% of current CPAs are nearing retirement. PwC planned to cut U.S. graduate hiring by a third over three years, according to an internal presentation obtained by Business Insider last August. KPMG has positioned its AI tools differently, describing them as a way to attract early-career professionals and redirect experienced auditors toward higher-risk judgment work. Scott Flynn, KPMG’s global head of audit, said the initiative is designed to help auditors “more effectively respond to risks and deliver deeper audit insights.”
KPMG’s latest Global AI Quarterly Pulse Survey found that 32% of organizations are deploying and scaling AI agents and another 27% are orchestrating multiple agents across their businesses. Three out of four global leaders said they would maintain AI investment priorities even amid economic uncertainty. The next-generation Clara agents for controls testing and financial statement analysis are expected before the end of the summer.
