McKinsey: for a few tokens more…

In 2025, McKinsey & Co. reached a symbolic milestone. It claimed to have used over 100 billion tokens via ChatGPT Enterprise. OpenAI was keen to mark this milestone by honouring the firm at a client event, highlighting one of the most widespread uses of generative AI in the corporate sector. Aside from this technological feat, one question remains: what is the value of AI-assisted advice?

by Antoine Boitez

IN LATE OCTOBER 2025, professional networks were all abuzz with an unobtrusive yet telling announcement. The US start-up OpenAI gave an award to McKinsey for topping the 100 billion token milestone. A huge figure, representing a volume claimed but not publicly audited. The equivalent of billions of words processed by AI. Taking an estimate of 0.75 words per token on average, this amounts to nearly 75 billion words generated, analysed or rephrased.

Humans still set the direction, but machines provide the momentum

Based on the firm’s 40,000-strong workforce working on 200-odd business days, this equates to an average use of just over 12,000 tokens per person per day. These are estimates, not precise figures, but they still indicate the extent to which AI has become part of the consultancy firm’s day-to-day work. For the partners, who are more closely involved in the scoping and reporting phases, it stands to reason that the use of AI is more widespread. Enough to transform working routines, at any rate.

100 BILLION TOKENS, AND US AND US AND US…

Let’s break it down according to the figures. On average, one ChatGPT token corresponds to three-quarters of a word in English. Multiply that by tens of billions, and you end up with a gigantic collection of documents: internal memos, sector overviews, data analyses, presentation materials – you name it. While the cost is unclear, estimates suggest a range of one to three million dollars a year for an Enterprise contract of this scale, based on standards observed with key accounts.

However, commentators and industry experts are playing it down. AI acts primarily as a co-pilot: it speeds up research, organises information, refines presentations and tests hypotheses. The firm’s flagship methodological frameworks, from the 7S Matrix to Porter’s Five Forces, remain largely driven by human expertise.

BARRIERS ARE COMING DOWN

While the announcement was not unanimously welcomed in the industry, criticism has been pouring in on Reddit and in certain professional communities: are clients paying high fees for deliverables partially generated by AI? McKinsey has rejected the notion of widespread replacement and stressed that AI supports upskilling and there are no plans for automated redundancies.

Nevertheless, doubts are beginning to arise. Are independent consultants and small firms, now equipped with comparable tools, threatening to undermine the monopoly of premium consultancy? AI is lowering the barriers to entry, but without bridging the gap in terms of experience, networks and credibility.

DIRECTION AND MOMENTUM

As 2026 approached, McKinsey epitomised a consultancy firm in transition. AI is now making capabilities that were once the preserve of giants accessible to everyone, while raising a key question for future clients: what return on investment can they expect? Should they pay for AI-enhanced expertise, or bring these tools in-house and take back control? Those more than 100 billion tokens mark our entry into a hybrid era. An era in which humans still set the direction, but machines provide the momentum. Who will actually set the course in the future is still unknown.