Real AI ROI Starts With Leaders: What McKinsey's New Research Means for You

McKinsey groups organizations into three stages of AI maturity, and the payoff grows at each one:

• Enablement: you give people general-purpose AI tools for individual tasks. Real enterprise value shows up for just 13% of leaders here.
• Automation: you rewire cross-functional workflows at scale. Value climbs to 24%.
• Reinvention: you redesign roles and operating models from the ground up to become one of the AI native organizations that pull ahead. Value reaches 48%.

The pattern is clear: the further you are willing to rewire the work, the more you get back. Yet almost nine in ten companies are still stuck in the first two stages.

That pattern makes sense. Real value shows up when leaders, and eventually their people, redesign the workflows they know best. A system does not redesign itself, and the people who can see what should change are the ones who understand both the work and the technology.

McKinsey says as much: choices like where AI creates value and how to run a mix of humans and agents do not bubble up from experimentation on their own. It is why the report calls leadership AI fluency the most significant upskilling effort in decades.

We see this every day. What leaders need most is a guide to help them work through where in their own work there is a real opportunity for workflow redesign, and then to do that work alongside them.

We call that skill Workflow Literacy, a term Stanford researchers introduced. It is the bridge between "I can use ChatGPT" and "I can redesign how my team operates," and our GED-RT framework helps leaders find the highest-leverage parts of their work to hand to AI first.

This process of rethinking where AI belongs in your work is what actually drives fluency. It is also why our fluency model is built around behavior change rather than just knowing about AI.

The instinct most leaders reach for is to "bring people along," and McKinsey is pointed that this means far more than nudging employees to try the tools. The real work is leading behavior change across the whole organization.

By the time a company automates, it has usually put tools in people's hands for individual tasks already. That is the easy part, and the study punctures the hope behind it: individual productivity with new tools does not predict organizational performance. Making one person faster does not, on its own, make the company better.

What the winners add is a way to pull ideas back up from the people doing the work, so the organization finds the higher-value opportunities no executive would spot from the top floor. That is the thinking behind the AI Implementation Sandwich we teach through our AI champion programs. McKinsey's reinvention data lands in the same place: innovation is championed from the top, then unleashed sideways, with people at every level surfacing ideas.

Our champions work this way on purpose, spread across every level of fluency rather than sitting in a technical elite, translating between what the business needs and what the tools can do.

The winners also make expectations concrete, with formal changes to workflows and real clarity on how people should work differently. Zapier does this with what its Chief People and AI Transformation Officer calls golden paths: documented, vetted ways of working with AI for a given function, from lead generation to a recruiting workflow. Clear paths give people the confidence to start.

A few moves follow from all of this:

• Treat AI as a change initiative, not a training rollout. Deploying the tools is the easy tenth of the work. Redesigning how the work gets done is the real job.
• Redesign one workflow, and start with your own. This is the 5.3x lever, so model it before you ask your team to.
• Build fluency across your team, not just at the top. Aim to get most of your people doing great work with AI. That is what company-wide AI training is built to deliver.
• Lead with trust, and build it by being honest. Trust was the one factor that predicted value at every stage. You build it by being clear about what you know and what you do not, following through, and investing in your people, never by promising nothing will change.

McKinsey's conclusion is that AI transformation is a story of human change, not technology deployment. I could not agree more.

It does not only start with leaders. It ends with them too, because the biggest driver of value in the most mature companies is a shift in how leaders run the place. The organizations pulling ahead are not winning because they bought better tools. They are winning because their leaders got fluent, redesigned the work, and earned the trust to bring their people with them.