Label the work before you hand it over

IBM's Institute for Business Value just surveyed 1,500 CHROs and senior workforce leaders across 21 geographies, plus 8,800 employees in 28 countries, about how AI is changing the work. The number I'd pin up first isn't about skills or culture. It's this: only 26% of organizations clearly define which work is human-led, which is AI-assisted, and which is AI-executed.

Three labels. Roughly one company in four has put them on its own work. Everyone else is running AI through processes where nobody has written down which parts a person owns and which parts the machine does.

The work didn't wait for the design

IBM's phrase for it is that work is changing faster than organizations are redesigning it. 52% of employees say the tasks they're assigned changed in the past year. So the work moved. The map didn't. What fills the gap is day-to-day improvisation — people quietly figuring out, one shift at a time, what to trust and what to check.

That improvisation has a cost, and the survey puts a number on it. 80% of CHROs say AI adoption creates "invisible" work: validating recommendations, fixing errors. 42% of employees report more workload or work nobody recognizes. The tool was sold as capacity. On the floor, a lot of it showed up as a second job — checking the first one — that isn't on anybody's list.

When it goes wrong, it lands on the person

Here's the part operators should read twice. 43% of employees say that when something goes wrong with AI, the blame falls on them. 41% of CHROs believe employees may not feel safe challenging or overriding what the AI produced. And 36% say unclear accountability is complicating deployment.

Put those together and you get the worst possible setup. The person is responsible for the output, doesn't feel safe overriding it, and was never told which steps were theirs to own. That isn't a people problem. It's a map nobody drew. In transportation and logistics, you would never hand a load to a driver without saying who's responsible for the pickup, the paperwork and the delivery window. Yet companies are handing work to a system with less clarity than a rate confirmation.

The label is the control

The companies that did the labelling report the payoff. IBM says organizations that clearly define workflows as human-led, AI-assisted or AI-executed report 18% risk reduction and 20% quality improvement. Those are self-reported, not audited — but look at what they come from. Not a better model. Not a bigger budget. A definition.

The confidence numbers point the same way. Where judgment is built into the work, 62% of CHROs see employee confidence in AI-enabled decisions rising. Where it isn't, 57% see it falling. Same workforce, same tools. The difference is whether someone decided, in writing, where the human stands.

This is what we do before we build anything. On the inbound inbox, the label was simple: the system drafts, a person approves, nothing reaches a customer without that person. Nobody had to wonder whose call it was. Reaction time still went from three-plus days to about a minute — because the check was designed in, not bolted on after something broke.

Do this before the next tool goes live

Take one process — the one AI is already touching. Write every step on its own line. Next to each, one label: human-led, AI-assisted, or AI-executed. Then add two more columns: who can override it, and who owns it when it's wrong.

You'll find three things fast. Steps nobody can label, because nobody ever decided. Steps labelled "AI-executed" that people are secretly checking anyway — that's your invisible work, now visible. And steps where the override column is blank — that's where the blame is going to land on whoever happened to be in the seat.

Three organizations in four haven't written down which work is whose. Label every step before you hand any of it over — the label costs an afternoon, and it's the cheapest control you'll ever build.

Sources: IBM, "New IBM CHRO Study: AI Puts Critical Thinking at the Center of Workforce Priorities" (press release, September 21, 2026; IBM Institute for Business Value with Oxford Economics, surveys of 1,500 CHROs and 8,800 employees, April–June 2026) · IBM Institute for Business Value, "2026 CHRO Study: Designing the Thinking Organization" (full findings) · HR Dive, "IBM report warns AI could erode human skills deemed vital by CHROs" (September 23, 2026)