Who Are the Big 4 AI Agents? Deloitte, EY, KPMG, PwC
The Big 4 AI agents are the agent platforms built by the four largest professional services firms: Deloitte Zora AI, EY.ai, KPMG Workbench, and PwC agent OS. They are real, they are in production, and they are sold as programmes through the firm rather than as software you operate yourself.
Here is what each one actually is.
The four platforms
Firm | Platform | Built with | Public scale claim |
|---|---|---|---|
Deloitte | Zora AI | NVIDIA | Agents across finance, tax, audit, and assurance |
EY | EY.ai | Microsoft and internal | 150 agents supporting 80,000 tax professionals, targeting 100,000 by 2028 |
KPMG | Workbench | Microsoft Azure AI Foundry | Around 50 agents connected, roughly 1,000 in development |
PwC | agent OS | Salesforce, CrewAI, AWS | Around 25,000 agents deployed across client operations |
Two things are worth noticing in that table. First, all four are built on someone else's model and infrastructure, so the firms are orchestrating rather than inventing. Second, the scale claims are about internal use and client delivery, not about agents running inside a client's own workflow with the client's own team approving the output.
The other meaning of the question
Some people asking this mean the four biggest AI model companies rather than the four biggest accounting firms. If that is your question, the usual four are OpenAI, Anthropic, Google, and Meta, with Microsoft and Amazon as the platforms they run on. These are model and tooling companies. They sell capability, not a role.
Neither reading describes something that will own a workflow for you. That is the part worth being clear about before you shortlist anything.
What they cost, as publicly reported
Ranges compiled from public reporting on Big Four AI engagements:
Engagement | Reported range |
|---|---|
AI readiness assessment | $25,000 to $75,000 |
Proof of concept | $50,000 to $250,000 |
Single use case deployment | $100,000 to $500,000 |
Platform rollout, first year | $500,000 to $3,000,000 |
Enterprise wide transformation | $500,000 to $5,000,000+ |
Treat those as indicative. They are published estimates rather than rate cards, and every one of them is scoped differently.
The structural point matters more than the numbers. Consulting still produces close to half of global revenue at EY, KPMG, and PwC, and closer to two thirds at Deloitte, so AI work is layered onto an existing billing model rather than replacing it. You are buying a programme with a partner on it.
Who should buy one, and who should not
Buy one if you are a large enterprise with a multi year transformation budget, a change management function, and a need for the firm's name on the risk. The governance scaffolding is genuinely the product.
Do not buy one if you have one workflow that is bleeding money and you want it fixed this quarter. A readiness assessment alone costs more than a full agent role deployed on your own systems. If what you actually want is a role off your team's plate, a programme is an expensive way to get there.
There is a real middle path. A Big Four platform can sit over a set of roles that are already running and producing numbers, rather than being the way you find out whether agents work at all.
Where this goes next
Should you hire another person or deploy an AI agent?
AI agents for accounting firms
Frequently asked questions
Who are the Big 4 firms? Deloitte, EY, KPMG, and PwC. The four largest professional services networks by revenue, traditionally audit and advisory.
Do the Big 4 sell AI agents? Yes. Each has a platform, and each sells agent work as part of advisory engagements rather than as a product you operate on your own.
What is the difference between Zora AI and ChatGPT? Zora AI is a platform of pre-built enterprise agents covering finance, tax, and audit functions, delivered by Deloitte on NVIDIA infrastructure. ChatGPT is a general purpose assistant that can be given tools. One is a programme, the other is a tool.
Are the Big 4 agents open source? No. All four are proprietary platforms, though EY.ai and KPMG Workbench are built on Microsoft infrastructure and PwC agent OS is built with Salesforce, CrewAI, and AWS.
Should a small business buy a Big 4 AI agent? Almost never. The entry point is a readiness assessment that costs more than a full year of running a single agent role on your own systems.
Is it worth paying for the brand? If the purchase needs board level sign off, regulated audit trails, and an outside name on the risk, yes. If the goal is recovered hours in one workflow, the brand is not doing the work.
