KPMG Put Claude in Client Work. CPA Firms Need a Plan

Key Takeaways

• Anthropic announced on May 19 that KPMG is embedding Claude inside Digital Gateway.
• The rollout gives Claude access to KPMG's global workforce of more than 276,000 employees.
• The important signal is platform integration, not chatbot access.
• Small and regional CPA firms should define a client-facing AI service statement now.
• The practical response is narrow: pick two workflows, document review controls and explain the value to clients.

KPMG did not just give employees another AI login. It put Claude inside Digital Gateway, the platform KPMG describes as central to client work.

Anthropic announced the alliance on May 19. The release says every one of KPMG's 276,000-plus employees globally will get access to Claude, and that Claude Cowork and Managed Agents are being embedded inside Digital Gateway. KPMG says the platform is where its tax expertise, proprietary tools and client data live together.

That is the part small firms should pay attention to. The Big Four is moving AI from side tool to delivery layer.

What exactly did KPMG announce?

Anthropic says KPMG is bringing Claude into its core business through a global alliance, starting with tools for tax and legal clients and expanding into private equity and cybersecurity.

The company claim is broad, so smaller firms should read it carefully. This is not independent proof that every KPMG engagement has transformed overnight. It is proof that a major professional-services firm is willing to position AI as part of client delivery, not just internal productivity.

That changes the sales conversation. A client comparing firms may now ask whether your firm can produce faster scenario analysis, cleaner advisory notes or better documented review workflows using AI.

Why does this matter for firms with 5 or 50 people?

Small firms do not compete with KPMG on global platform spend. They compete on trust, speed, specialization and partner access.

AI changes that mix. A regional firm that uses AI responsibly can look more capable than its size suggests. A firm that cannot explain its AI policy may look dated, even if its technical work is sound.

PwC's finance-agent benchmark shows the gap. PwC says 79% of executives report AI agents being adopted in their companies, but only 34% are using them in accounting and finance. That means many clients will hear enterprise AI claims before their accounting workflows are mature enough to evaluate them.

What should a small CPA firm say to clients?

Write a one-paragraph AI service statement before clients ask.

It can be plain: "We use approved AI tools to speed research, drafting, reconciliations and advisory preparation where appropriate. We do not let AI replace professional judgment, client confidentiality controls or partner review. Human reviewers remain responsible for final work product."

That statement does not oversell. It tells the client three things: the firm is not ignoring AI, client data still has boundaries, and a human remains accountable.

The internal version can be even shorter: "AI may help draft or organize work, but it does not approve client conclusions, override confidentiality rules or replace partner review." That sentence is useful because staff can remember it under client pressure.

What does platform integration change?

Platform integration changes the AI conversation from "which chatbot do you use?" to "where does AI sit in the delivery workflow?"

A browser tab can help an employee draft. A delivery platform can shape how tax research, client requests, workpapers, scenario analysis and review notes move through the firm. That is why the Digital Gateway detail matters. KPMG is not merely saying its people can ask Claude questions. It is saying Claude will operate inside the environment where client work is organized.

Small firms can borrow the logic without copying the infrastructure. A firm using Microsoft 365, Karbon, Canopy, QuickBooks, Xero or a document portal can still define where AI is allowed to touch the workflow. For example, AI may be approved for summarizing a client email thread, drafting a variance memo or preparing a tax-research outline. It may be banned from unapproved client-data uploads, final conclusions or unsupervised client communications.

The point is to make AI placement deliberate. If the firm does not decide where AI belongs, staff will decide one prompt at a time.

Which workflows should firms pilot first?

Pick two workflows that are visible enough to matter but controlled enough to review.

• Client advisory prep: Use AI to draft first-pass explanations after staff identify the data, issue and recommendation.
• Month-end variance narrative: Use AI to improve clarity after staff document the account movement and supporting evidence.
• Tax research summary: Use AI to summarize source material, with a manager checking citations and final position language.
• Internal workpaper cleanup: Use AI to improve wording, not to invent procedures or conclusions.

The rule is simple: AI can help package reasoning. It should not replace the reasoning.

What are the three moves available now?

First, assess your current AI tool stack. Know which staff use ChatGPT, Claude, Copilot, QuickBooks AI or other tools. Check whether firm data is allowed, restricted or banned in each tool.

Second, define your client-facing AI service statement. Put it in onboarding, proposals and staff talking points. Clients should hear a consistent answer.

Third, pick one workflow to demonstrate. A firm does not need a platform like Digital Gateway to show that AI improves a deliverable. It needs a before-and-after example with human review clearly marked.

A practical demo could be a monthly close package where the staff accountant prepares the account movement, AI drafts the plain-English client note, and the manager signs off on the final explanation. Another could be a tax planning memo where AI summarizes the source material but the partner writes the conclusion. A third could be a client onboarding checklist where AI helps turn prior-year records into a cleanup plan, with no confidential data pasted into an unapproved tool.

Those examples are smaller than KPMG's announcement. That is the point. Clients do not need every firm to look like KPMG. They need their accountant to show a controlled way to use modern tools without weakening confidentiality, evidence or judgment.

Nexairi's AI operational standards guide for accounting firms covers the governance layer. This KPMG move adds the positioning layer: how the firm explains its AI capability in the market.

The small-firm advantage is not scale

KPMG's advantage is infrastructure. A smaller firm's advantage is proximity. Partners can choose a narrow use case, teach the team, inspect the work and explain the control to clients without waiting for a global rollout. The firms that win will not pretend to be Big Four. They will be specific about where AI makes their service better and where human judgment stays in charge.

Frequently Asked Questions

What did KPMG actually announce about Claude?

Anthropic announced on May 19 that KPMG is embedding Claude inside Digital Gateway, the platform KPMG uses for client work. The rollout gives Claude access to more than 276,000 employees globally. This is platform integration, not just chatbot access. Claude will operate where client tax research, workpapers and advisory work are organized.

Does a small CPA firm need to match what KPMG is doing?

No. Small firms compete on trust, speed and partner access, not global platform spend. A regional firm that uses AI responsibly can look more capable than its size suggests. The goal is to define where AI belongs in your workflow before staff decide one prompt at a time.

What should a CPA firm say to clients about AI?

Write a one-paragraph service statement before clients ask. Cover three points: the firm uses approved AI tools for research and drafting, client data stays within confidentiality controls, and a human reviewer remains accountable for the final work product. Staff should be able to repeat that answer consistently.

Which workflows should small firms pilot with AI first?

Pick two controlled workflows. Good starting points are client advisory prep (AI drafts after staff identify the issue), month-end variance narratives (AI improves clarity after staff document the movement) and tax research summaries (AI summarizes sources while a manager checks citations). The rule: AI can help package reasoning, not replace it.

Sources

• Anthropic - KPMG integrates Claude across its core business
• PwC - AI agents for finance

Related Articles on Nexairi

• Accounting Firms Need AI Standards Before AI Spreads
• What Accounting Staff Should Never Paste Into ChatGPT
• Microsoft Copilot Can Now Reconcile Accounts and Explain Variances

What should partners do before the next proposal?

Before the next proposal, partners should write down three things: what AI can do in the firm, what it cannot do, and where a human review is required.

That becomes useful sales language because it is specific. A firm can say it uses AI to prepare faster research summaries, improve advisory drafts and organize client records, while final positions remain reviewed by a qualified professional. That is stronger than saying "we use AI" and safer than implying the tool performs professional services by itself.

Use 2026 to define the policy and 2027 to measure whether it changed service delivery. If the firm cannot show one better client deliverable by then, the AI policy is only a memo.

The firm should also train staff on the answer. If a client asks a senior associate whether the firm uses AI, the answer should not depend on which person happens to be on the call.