The fact: On 19 May, Anthropic and KPMG — one of the "Big Four" global consulting firms — announced a strategic alliance. Claude is now deployed across all 276,000 KPMG employees worldwide, integrated into internal tools, audit methodologies and client deliverables. This is no cosmetic technology partnership: it's an operational shift across the entire business.

The Codito analysis: KPMG isn't a software vendor dabbling with a model for fun. It's a firm billed by the consulting hour, with some of the strictest compliance standards in the world. If it's rolling Claude out to 100 % of its teams, it's because it has done the maths: (a) the productivity ROI far outweighs the cost of integration, (b) the reputational risk of waiting is greater than the risk of moving, (c) not doing it means losing bids to competitors who have. That calculation holds for many other professions: lawyers, accountants, architects, engineers, recruiters, marketers. Wherever the product is a structured deliverable, the KPMG pattern applies.

What it means for you: If your clients are large corporations — the ones buying from KPMG, Deloitte, EY or PwC — their deliverables will progressively be produced by humans assisted by agents. You will be pulled towards the same cadence. If your clients are SMEs, the pressure will come from their competitors who are already mobilising AI. Whatever your market, productivity per head is once again the referee.

The number of the week 276,000. That's the number of KPMG employees who now have Claude access built into their methodology. For comparison: France's active workforce is around 30 million. In a single announcement, KPMG just shifted the equivalent of nearly 1% of the French active workforce to a new way of producing consulting.

The KPMG announcement echoes the same question we hear in every company we meet : " We can see things are moving. But concretely, where do we start ? "

Here is the KPMG method, broken down into four steps — and how to replicate them in an SME, without 276,000 employees.

Step 1 — Scoping: which business problem are you addressing, exactly ?

KPMG didn't deploy Claude " to do AI ". The alliance targets precise use cases : account analysis, first-draft deliverable generation, audit meeting summaries, automation of everyday advisory work. Each case has a measurable before/after. In an SME, transpose this : identify two or three recurring tasks — sales qualification, drafting specifications, client summaries — where you could quantify the time saved. Without this scoping, AI remains a collective gadget.

Step 2 — Choosing the model: not one single model, the right model for each case.

KPMG didn't commit to Claude by chance : Anthropic is positioned for the enterprise (security, confidentiality, managed agents). But internally, several models coexist depending on the use case. In an SME, the same principle holds : don't lock in a single vendor. Evaluate Claude, ChatGPT, Mistral and Gemini on your real cases, choose case by case, and reassess every quarter. The market moves faster than your annual contracts.

Step 3 — Rolling out in waves: not everyone at once.

KPMG is deploying by cohort — not 276,000 people overnight. First the support functions, then the most AI-mature teams, then everyone else. In an SME, the logic is identical : start with the employees who are already comfortable, measure, formalise the best practices, expand. A week's delay on one wave costs less than a month of failed adoption at scale.

Step 4 — Governance: who decides what, who measures what.

KPMG published an internal usage charter for its 276,000 employees : which data can be submitted to AI, which deliverables must be validated by a human, what traceability is required. In an SME, it's one A4 page. But it exists. Without it, individual usage takes over — and no one knows what's happening with your data and your deliverables anymore.

One point that runs through all four steps: the quality of your documentation. When KPMG deploys Claude, it's on top of processes already mapped, business rules already written down, data already structured. AI accelerates what is explicit — it doesn't guess what isn't. If the way you qualify a prospect, draft a quote or analyse a file lives only in your employees' heads, no AI — be it Claude, GPT-5 or Gemini — will be able to reproduce it faithfully. Garbage in, garbage out remains the exact adage. Document your processes, your criteria, your edge cases precisely: that is the raw material the AI will work with. Without it, you'll get a lukewarm average instead of your own expertise.

You don't have 276,000 employees. You don't need 276,000. But you do need the same methodological rigour — and documentation to match. That is precisely what we build with our clients through our AI Audit and our Codito AI Partnership.

Book a strategy call

Discover our Codito offers

To close this edition, a fresh look behind the scenes.

This week, a French engineering firm confirmed the signing of a development project with us: the design and deployment of an AI agents platform dedicated to prospecting and quote generation.

Concretely, it means replacing a chain that is manual today — prospect research, qualification, drafting the quote from a set of specifications — with an orchestration of agents that :

• Identify high-potential prospects based on specific business criteria
• Automatically qualify each opportunity before it takes up a salesperson's time
• Generate a first priced version of the quote, ready to be validated by a human

The stakes go beyond saving time. It's about enabling sales teams that used to handle ten files a week to handle thirty — with no drop in quality. If anything, with higher quality, because each qualification draws on a more exhaustive analysis than a rushed human could produce.

At our scale, this project is exactly the same kind of move KPMG has just made across 276,000 employees. The pattern is spreading — well beyond the Big Four. And that is precisely what we help our clients do.