# Claude Enters the Big Four: When Consulting Stops Selling AI and Starts Operating with AI

> Big Four firms are shifting from advising on AI to embedding AI agents in their delivery and client platforms.

- Author: Viktor Berthelius (BRTHLS)
- Published: 2026-05-14
- Updated: 2026-06-29
- Category: ai operating models
- Tags: claude, big-four, ai-operating-model
- Language: en
- Canonical: https://www.brthls.com/magazine/claude-big-four-operating-ai-en
- Source: BRTHLS Magazine — https://www.brthls.com

---

## Problem

For a long time, the major consultancies have sold digital transformation from the outside: diagnosis, roadmap, governance, vendors, training and deployment. Agentic AI changes that position. When a firm like PwC deploys Claude Code and Claude Cowork in its own workforce, and KPMG integrates Claude into its client platform, the consultancy stops being just an advisor.

It starts to become an operator.

## Thesis

The May 2026 news isn’t that the Big Four use AI. That was inevitable. The news is that they are turning AI into an internal delivery capability: engineering, deals, tax, legal, finance, cybersecurity and proprietary platforms.

When consulting operates with agents, the client doesn’t buy recommendations. They buy a packaged execution model.

## Framework

This shift has four layers:

- **Workforce enablement:** hundreds of thousands of professionals with access to models and agents.
- **Delivery system:** agents integrated into internal software, not just separate chats.
- **Client platform:** products and processes delivered on that infrastructure.
- **Governance layer:** trust, compliance, auditability and decision rights as part of the service.

Mini‑case: a mid‑size company hires an AI audit. Previously it received slides, benchmarks and a plan. Now it can also receive an operating system: configured agents, connected workflows, controls, training and a delivery layer that the consultancy already uses in its own house.

**Measurable signal:** percentage of consulting deliverables that end up as production workflows, not as recommendations.

**Position:** a consultancy that does not operate with AI will have decreasing authority to sell AI.

**Breath:** advising on agents without working with agents will be a hard position to defend.

## What changes for the client

The buyer should no longer ask only:

- what experience the firm has in AI
- what use cases it has seen
- what methodology it proposes

They should ask:

- which part of its own operation already runs with agents
- what internal controls it uses
- what data, logs and thresholds it can show
- what happens when an agent fails
- what ownership remains with the client at the end

The difference between selling AI and operating AI lies in the evidence.

## Common mistake

The counter‑example is assuming that a large partnership solves the client’s operating model. It does not. It can speed deployment, but it can also create dependency if the client does not understand which decisions stay internal, which workflows depend on the consultancy, and which capabilities must be absorbed.

The risk is not hiring help. The risk is outsourcing judgment.

## Protocol (3 steps)

1. **Evaluate the consultancy on real operation.** Which agents it uses, where, with what controls and with what results.
2. **Require system transfer.** Playbooks, owners, logs, closure criteria and rollback mechanisms.
3. **Separate capacity from dependency.** The firm can accelerate, but the operating model must be governed by the business.

| Question | Why it matters |
| --- | --- |
| Which agents are used internally | Shows real maturity |
| Where the controls are | Separates demo from production |
| What remains in the client’s hands | Avoids dependency |
| What is measured afterwards | Turns delivery into learning |

## Related

- [AI Operating Models in 2026: the 5 patterns that do scale](/magazine/ai-operating-models-2026-5-patrones-es)
- [AI Governance Backlog: turning risk into executable work](/magazine/ai-governance-backlog-risk-to-work-en)
- [Executive Review Stack for AI: what a CEO should look at each week to govern without theater](/magazine/executive-review-stack-for-ai-the-weekly-ceo-stack-for-real-governance)

## Sources consulted

- [PwC is deploying Claude to build technology, execute deals, and reinvent enterprise functions for clients](https://www.anthropic.com/news/pwc-expanded-partnership)
- [KPMG integrates Claude across its core business and workforce of more than 276,000](https://www.anthropic.com/news/anthropic-kpmg?939688b5_page=1)

## Next step

If you are hiring AI help, don’t ask only about strategy. Ask what operating system they will leave running when they leave. We can review it in a [diagnostic](/en/contact).

---

*Translated from the Spanish original with AI assistance and reviewed for accuracy. [Read the original in Spanish](/magazine/claude-big-four-consultoria-vende-ia-opera-ia-es).*

---

_Cite as: Berthelius, V. (2026). "Claude Enters the Big Four: When Consulting Stops Selling AI and Starts Operating with AI". BRTHLS Magazine. https://www.brthls.com/magazine/claude-big-four-operating-ai-en_