Anumber from Deloitte's 2026 Global Human Capital Trends survey, covering more than 9,000 leaders across 89 countries: 6 percent say they are making progress in designing how humans and AI actually work together. Not finished. Making progress.
Organizations taking a technology-first approach are 1.6 times more likely to miss their return expectations compared to those that redesign roles and workflows around human-AI collaboration.
BCG's 10/20/70 framing puts it starkly: 70 percent of the investment that drives AI value is in people and processes, not technology. Only 25 percent of companies report meaningful ROI. The gap between the 25 percent and the rest comes down to how they reorganized work around the technology.
What does the other side actually look like in practice?
One of the few concrete examples in the public record comes from McKinsey's documentation of a consumer brand that redesigned its creative production process. The old workflow involved months of iteration across internal teams and external agencies, with the familiar cycle of feedback, revision, and rework. The organization took the entire process apart. They decomposed it into hundreds of microtasks: concept generation, pretesting with focus groups, risk assessment, content versioning, agency coordination. Then came the actual design act. Each microtask was evaluated independently for whether it should be handled by a human, an agent, or some combination. What emerged was a different process altogether, with different handoffs, different decision points, and roles that hadn't existed before.
That decomposition is what separates the 6 percent from the rest. Most organizations are mapping AI onto existing workflows and wondering why the gains plateau. The workflow itself is the ceiling. The organizations pulling ahead are rethinking what McKinsey calls "decision rights." When agents can surface competitive pricing across thousands of sources in hours instead of weeks, the operational pressure lands downstream: who acts on it, and how quickly? That's organizational design. It touches team structure, reporting lines, how performance is measured, what meetings exist and why. New roles are already emerging around this: people who design agent workflows, people who lead blended human-agent teams, people whose job is figuring out which decisions still need a person in the room.
What's striking about the organizations doing this work is how ordinary it sounds when they describe it. Someone sat down and asked what the work actually is before deciding which parts to automate. They mapped the microtasks. They assigned decision rights deliberately. None of it is glamorous. But it produces a kind of organizational clarity that compounds, because every subsequent agent deployment fits into a structure that was designed to receive it.