The most revealing thing about KPMG's push into artificial intelligence is what the firm does not claim to have invented. Microsoft supplies cloud and copilots. Google supplies models and infrastructure. ServiceNow supplies workflows. Anthropic supplies Claude. KPMG supplies the part that becomes painfully important after the demo: an answer to who approved the system, which evidence it used, where it can act and what happens when it is wrong.
That proposition comes naturally to a business descended from accountants. KPMG LLP, the U.S. audit, tax and advisory partnership, lives around consequential decisions. A public-company audit supports confidence in financial statements. A tax opinion interprets rules whose details can move millions of dollars. A cyber or supply-chain project changes systems that a large organization cannot simply switch off. The work is rarely glamorous. Its value appears when executives, boards, investors and regulators can proceed without guessing.
KPMG calls itself a professional-services firm, a description broad enough to hide the machinery. It sells expert labor, certainly, but also methodologies, proprietary platforms, implementation teams and ongoing operations. It helps finance chiefs close books, tax departments manage filings, boards examine risk, private-equity owners improve portfolio companies and government agencies modernize old technology. Increasingly, the product is not a report delivered at the end. It is a controlled system that keeps running.
A firm that sells permission to proceed
The U.S. story begins in New York in 1897, when Scottish immigrants James Marwick and Roger Mitchell formed Marwick, Mitchell & Company. Their opportunity was independent review at a time when American business was growing larger and its outsiders needed reliable numbers. The modern initials arrived much later. KPMG joins the names Klynveld, Peat, Marwick and Goerdeler, lineages brought together when Peat Marwick International and Klynveld Main Goerdeler merged in 1987.
That history matters because KPMG still occupies two roles that tug in different directions. As an auditor, it must be independent, skeptical and willing to disappoint the company paying the fee. As an adviser, it is hired to help that company move faster. Independence rules mean some consulting services are impermissible for audit clients. The limitation is commercially awkward, but it also teaches the firm to separate claims from evidence and enthusiasm from authorization. In the AI market, those habits look less like old accounting baggage and more like product requirements.
AssuranceConfidence in statements, systems and controls
Three doors, one client problem: understand what is true, what is allowed and what to do next.
From hours and projects to systems that stay
KPMG's traditional business model is straightforward: a partnership assembles credentialed teams and charges fees for engagements. Audits recur. Tax and consulting work can be project-based. Complex deals and regulatory changes create bursts of demand. The global KPMG organization reported $39.8 billion in fiscal 2025 revenue, but that is network revenue, not a disclosed figure for the legally separate U.S. partnership. Supplied company data puts the American workforce at roughly 36,000 people.
Managed services bend that model. KPMG can run a defined slice of finance, tax, compliance, legal contracting, learning or customer operations under a multi-year, as-a-service arrangement. The client receives a predictable cost and access to specialized people and technology. KPMG receives recurring work and a closer view of the process. Advice becomes operations; operations generate data; data reveals the next improvement. That loop is harder for a client to replace than a presentation.
How an engagement becomes an operating layer
The alliance stack is the product roadmap
Large clients already have software allegiances, technical debt and thousands of workers inside Microsoft, Google, Oracle, SAP and ServiceNow environments. KPMG does not need to persuade them to abandon those systems. It places accountants, engineers, industry specialists and change teams on top of them. This makes the firm's apparent weakness - not owning the foundational cloud or model - a useful neutrality. It can meet a client in the stack the client already has.
The investments are substantial. In 2023, KPMG announced at least $2 billion over five years around its Microsoft cloud and AI alliance. In 2024, it committed $100 million to its U.S. Google Cloud practice and estimated the work could add $1 billion in growth. Its ServiceNow relationship includes a multi-year expansion and $40 million in KPMG services over three years. The Oracle alliance stretches back more than three decades.
Then, in May 2026, KPMG and Anthropic announced Digital Gateway Powered by Claude. Digital Gateway is KPMG's Azure-based environment for tax knowledge, proprietary tools and client data. Putting Claude inside it allows professionals and clients to assemble agents without carrying sensitive work across a collection of chat windows. KPMG said an agent for changing tax rules that once took weeks could be built in minutes. The alliance also extends Claude access to the network's 276,000-plus global workforce and targets private-equity portfolio companies.
“Our clients depend on us where accuracy, judgment and knowledge matter most.”Tim Walsh, Chair and CEO, KPMG US
Trust is useful only when it changes the workflow
Every consulting firm now speaks about responsible AI. KPMG's sharper opportunity is to make the phrase concrete. Its Clara audit platform uses AI agents and continuous analysis to focus auditors on unusual risks and complex judgment. Its AI assurance work examines whether systems are governed as promised. Its Trusted AI framework pushes security, privacy, fairness, explainability and accountability into design and operation. None of this makes a model infallible. It makes failure easier to anticipate, contain and explain.
That is valuable to customers whose cost of error is asymmetric. A retailer can tolerate a mediocre product description; a bank cannot casually tolerate discriminatory credit logic. A marketing team can discard a bad draft; a tax department needs to know which rule produced an answer. A software prototype can break; a public-company control cannot quietly improvise. KPMG fits where experimentation crosses into an obligation.
Who hires KPMG - and why
The customer list spans public and private companies, financial institutions, private-equity firms, healthcare systems, governments, universities and nonprofits. Inside them, the buyer may be a chief financial officer facing a reporting deadline, a tax leader confronting new cross-border rules, a chief information officer replacing an aging platform, an audit committee asking what could go wrong or an operations leader trying to take cost out of a supply chain.
Their shared problem is complexity with witnesses. Many people, systems and regulators will inspect the result. KPMG differentiates itself through the combination of a regulated assurance practice, deep functional teams, industry organization and technology alliances. The firm says it was the first Big Four member to organize itself along the same industry lines as clients. That structure matters because a control, tax treatment or cloud migration behaves differently in banking, healthcare, manufacturing and government.
Use KPMG to verify
Financial statements, controls, technology, sustainability reporting and AI governance.
Use KPMG to navigate
Tax obligations, regulation, risk, transactions and board-level uncertainty.
Use KPMG to change
Finance, supply chain, customer operations, data estates and cloud platforms.
Use KPMG to run
Defined business processes under managed, multi-year service arrangements.
A crowded market with four familiar names
KPMG's closest competitors are Deloitte, PwC and EY. All four can bring auditors, tax specialists, consultants and global reach. Deloitte is larger globally and has a vast consulting operation. Accenture and IBM compete hard in implementation and managed technology. McKinsey, BCG and Bain meet KPMG in strategy. Grant Thornton, RSM and specialist firms can be more focused or less expensive. Clients can also build internal teams, especially when the expertise is needed every day.
KPMG's pitch is strongest when the assignment joins several disciplines: a cloud implementation with controls, a deal with tax consequences, AI adoption inside a regulated function or an operating redesign that must satisfy an audit committee. Its pitch is weaker when a buyer wants only commodity labor, a single software product or unrestricted consulting from its statutory auditor. The firm is not an all-purpose answer. It is an integrator of judgment where functions collide.
The human system behind the control system
Professional services scale through apprenticeship. KPMG's public values - integrity, excellence, courage, together and “for better” - read like behavioral controls for a business whose output is often a judgment. Its Lakehouse learning center in Orlando makes that system physical: professionals gather to train, compare methods and absorb the firm's way of working. Community Impact Day, responsible-AI commitments and education programs add a civic layer to the employer story.
Culture claims deserve the same skepticism as any corporate claim. Long hours, deadline pressure and hierarchical advancement are familiar across the Big Four model. Yet training is not decoration here. If KPMG wants thousands of people to use agents without outsourcing their judgment, it must teach them when to challenge a result, document a decision and stop a workflow. The AI transformation is partly software and mostly management.
KPMG began by helping outsiders believe a set of accounts. The object requiring belief has expanded: a tax engine, a cloud migration, a supply chain, an AI agent. The firm's market position is neither pure accountant nor pure technologist. It is the party hired to connect expertise, software and evidence so that a consequential system can be used with eyes open.
That leaves KPMG with a practical test. Enterprise AI will produce no shortage of fluent answers. Customers will pay for the ones that survive a harder set of questions: Where did this come from? Who checked it? What is it allowed to do? And can we show our work?