The 40,000 Humans, 20,000 Agents Problem

The CFO vs CIO Problem Is Actually an Org Design Problem

One of the sharpest observations in the interview was about a conversation Sternfels keeps having with CEOs. The CFO says: we are spending money on AI but not seeing enterprise-level returns. Why not be a fast follower? The CIO says: if we are not in the lead we will be disrupted.

Both are right and both are missing the point.

McKinsey's finding, after studying hundreds of implementations, is that more than half of what makes AI adoption work is organizational change, not technology. You can deploy the best tools in the world and still fail because your approval chains are too long, your middle management is too thick, or your departments are too siloed to let a single AI-enabled workflow touch more than one of them.

He used a mortgage as the example. Origination. Credit scoring. Collections. Servicing. Four or five departments, all in sequence, all with walls between them. If a single intelligent process can handle the workflow end to end, why do you still need four departments? You do not. But getting a large organization to remove those walls is a political and cultural project, not a software one.

This applies directly to professional services firms in Kenya and across Africa. The firms that will move fastest are not the ones that buy the most sophisticated tools. They are the ones that redesign their operations around what those tools actually enable.

What Clients Will Pay For Next

Sternfels was asked a pointed question: if AI can commoditize the analysis and insight McKinsey has long provided, what will clients pay for?

His answer: they will pay to double their market cap. There will always be a harder problem. The work a McKinsey consultant did in 1993, clients do themselves today. The work they do in 2025, clients will do themselves by 2030. The job is to stay ahead of that curve.

For smaller consultants and boutique firms, this is both a threat and the clearest possible brief. If your value proposition is expertise that can be replicated by a well-prompted AI, you have a window to fix it. If your value proposition is judgment, relationships, the ability to ask the question nobody thought to ask, you are more durable than you think.

The Three Skills McKinsey Is Looking For Now

After analysing 20 years of internal data, McKinsey identified three things that actually predict success at the firm, none of which are what they used to hire for.

Resilience. The candidate who had a setback and recovered was more likely to make partner than the one with perfect marks. Failure survived is a credential.

Real experience working with others. Team sports. Retail jobs. Anything that forced you to collaborate under pressure with people who were not your friends.

Aptitude to learn new things over mastery of existing ones. They now deliberately create assessment environments where no candidate has any pattern recognition, just to see how people navigate genuine uncertainty.

Beyond those, they are exploring three more for the post-AI era: the ability to set aspiration (models do not aspire), judgment (models do not have it), and discontinuous thinking (models are excellent at the next logical step but poor at the creative leap). Hence a renewed interest in liberal arts graduates.

This is worth reading carefully if you hire people or if you are thinking about your own career. The skills that got you here are not the ones that carry you forward. The firms that figure out how to recruit for judgment and creative discontinuity before their competitors do will have a durable advantage.

The Accountability Question

Sternfels was asked directly about OxyContin, bribery charges in South Africa, and conflict of interest accusations. He did not dodge it.

His framing was useful: some of this fell in the humble camp, where they got it wrong and had to change. Some fell in the courageous camp, where they were criticized for things they believed were right and pushed back anyway.

The lesson from the humble camp: good intentions and a good ethos are not a compliance framework. McKinsey spent a billion dollars modernizing its risk and compliance infrastructure, brought in the head of internal audit from Apple and the head of compliance from Walmart. Not because the partners were bad people. Because as an organization scales, values without systems become aspirations without enforcement.

This is the accountability paradox for professional partnerships. The ethos of the firm was never to grow at all cost. But without compliance architecture, you cannot guarantee that the ethos survives contact with 40,000 people operating across 65 countries.

The Outcomes Question

The interview ended with where McKinsey wants to be in ten years: known not as a great adviser but as an impact partner. A third of revenues are already tied to outcomes, not advice. Sternfels wants that to be a majority before he finishes his tenure.

This is the direction the whole industry is moving. Clients are increasingly unwilling to pay for recommendations they then have to implement themselves. They want partners who share the risk and stay until the thing actually works.

For smaller firms and solo consultants, this is operationally hard to do and strategically unavoidable to consider. What would it look like to underwrite an outcome for a client rather than sell them a plan? What would you need to believe about your own capability to make that offer?

McKinsey turning 100 is a reminder that the firms built to last are the ones willing to keep asking what they are actually for.