Why McKinsey Can't Make You AI-Native (And What Can)

Every company that wants to become AI-native is still using the same method to get there: hire consultants, run workshops, write a roadmap, and hope the organisation changes. That model was built for a slower world.

McKinsey and its peers were designed for narrow, well-scoped problems. A market entry assessment. A cost-reduction programme. A supply chain review. These are problems where sampling works: a small team can interview enough stakeholders, synthesise the findings, and produce something useful without understanding every corner of the organisation.

AI transformation is not a narrow problem. It touches every function, every workflow, every team, every tool, and almost every role. You cannot sample your way to a transformation. You cannot produce a roadmap that addresses something this broad from the outside. And you cannot hand the roadmap to the company and expect it to execute. That is not how transformation happens.

What Consulting Was Built For

To be fair to the model before criticising it: consulting firms built their approach to solve a specific problem. Organisations have questions they cannot answer from the inside, usually because they lack the expertise or the outside perspective. A new market. A regulatory change. A merger integration. The consultant brings the expertise in, spends a defined period inside the organisation, and delivers an answer.

This works when the problem is bounded. You can scope it. You can staff it. You can deliver a report and it answers the question.

None of these are wrong. The problem is applying this model to something it was not designed for.

Why AI Transformation Is Different

Raphaël Dabadie worked on this question at YC (P26) and articulated it precisely: "AI transformation touches every function, every workflow, every team, every tool, and almost every role inside the company. When led internally, the transformation does not get any simpler. A small group of key people suddenly has to understand how the entire company works, decide where AI should be integrated, convince teams to change, and keep updating the plan as models improve. That is an impossible amount of context to hold manually."

The consulting model samples. It selects stakeholders to interview, workflows to map, teams to assess. What it misses is how the company actually works. Not the stated processes. The real ones. The exceptions everyone knows but nobody documented. The informal routing paths. The things that happen between the documented steps.

Dabadie again: "This is why so many enterprise AI deployments fail: they are built on the official company instead of the real one."

"AI adoption adds tools to the existing company. AI transformation redesigns the company around what AI now makes possible."

That distinction is the gap the roadmap cannot close. Adoption is additive. Transformation is structural. A consultant can advise on the former. Only someone inside the real systems can drive the latter.

The Method Has to Match the Problem

The argument is not that McKinsey is bad. The argument is that the method does not match this problem.

Dabadie: "If AI is the transformation, AI also has to be the method."

This sounds abstract. It is actually practical. It means:

• The transformation cannot be managed through a roadmap delivered every quarter. It has to run continuously, updating as models improve and as agents surface new capabilities.
• The transformation cannot be understood from outside the company. Someone has to be inside the live systems, watching what the agents encounter, refining the rules, fixing what breaks.
• The transformation cannot be handed off to a team to execute. Whoever is building the AI systems has to understand the company well enough to build them correctly. Not from documentation. From being inside the operations.

This describes an embedded operator, not a consulting engagement. The difference is not just duration or cost. It is accountability. The consultant delivers the roadmap and exits. The operator owns what runs.

The Companies That Are Actually Transforming

YC rebuilt their knowledge and office hours systems by embedding engineers who watched the agents fail and fixed them overnight. Not by delivering a roadmap about what to build.

Airtable's CEO Howie Liu spent a year embedded in his own company's workflows, redesigning them around AI. Not commissioned from outside. Run from the inside. The restructuring happened because the person accountable for the outcome was also inside the live systems.

Anthropic writes more than 90% of its production code with AI. Not because they hired consultants to advise on AI strategy. Because they are the people building the systems and they built their own company around them.

The pattern is consistent. Transformation happens from inside, by people who are accountable for what runs, not for what they recommended.

What You Actually Need

The gap that consulting cannot fill is the one between the roadmap and production. Between knowing what to do and having something running. Between the discovery phase and the live system.

What gets AI into production is an operator who arrives on day one, accesses the live systems, understands how the company actually works, ships the first agent in two weeks, and stays until the company can own it.

That is the nativefirst model. Not a roadmap. Not a workshop. A production system, shipped in weeks, built from the real company, accountable to what runs.