In Adecco’s seventh Global Workforce of the Future report, published Nov. 19, the staffing and workforce-services company points to a problem many employers would rather treat as a temporary rollout issue: more workers say they are using AI agents, but fewer say they are being trained to use them. Among 40,500 surveyed workers across 36 markets and 21 industries, 74% reported using AI agents at work, up from 64% in 2025. Over the same period, the share reporting AI training fell to 67% from 77%.
That matters because the business question is no longer whether companies can switch AI tools on. It is whether they are redesigning jobs, decision rules, and management practices fast enough for those tools to produce durable productivity rather than a brief burst of novelty, confusion, or extra workload.
Adecco’s data does not settle that question on its own. The report is based on a self-administered online survey fielded in June and July 2026, not an audit of employer systems or an objective productivity study. But it does capture a practical implementation gap that software dashboards often miss: adoption can rise even while readiness falls.
Adoption is getting ahead of preparation
The headline numbers are stark because they run in opposite directions. More respondents say they are using AI agents; fewer say they are receiving training. Adecco does not show whether that training decline reflects less formal instruction, more informal learning, a change in question wording, or a shift in the respondent mix. Still, the gap is important because AI use without preparation changes the risk profile of deployment.
The report says 66% of workers believe AI has changed the skills required for their jobs. Among surveyed AI users, 93% of those who completed training said AI made them more productive, compared with 47% of those who had not completed training. That is not proof that training caused the difference. People who receive better management may also be more likely to receive training and to report positive outcomes. But the size of the gap is hard for employers to ignore.
The same pattern appears in workflow design. Workers who said AI was well integrated into workflows reported saving 20% more time per day than those experiencing lower integration, and they were seven times less likely to feel overwhelmed by AI tools. That finding gets at a point many enterprise AI programs still underplay: an agent does not create business value just by completing a task. The surrounding work has to change too.
If an agent drafts an email, pulls information from files, or handles a routine service step, someone still needs to decide what data it can touch, when its output requires review, how exceptions get escalated, and what happens to the time it appears to save. Without those operating decisions, “time saved” can quickly turn into “more tasks assigned.”
The management problem is bigger than the software problem
Adecco’s leadership findings are even more revealing than the adoption numbers. Only 19% of respondents said leaders are empathetic to the effects of change. Just 21% said their concerns about new technologies are taken seriously, and 39% said they feel supported by their manager to adapt.
Those are weak numbers for any workplace transformation, but they matter especially in AI rollouts because the technology often changes how work is judged before it clearly changes how work is structured. Workers can be asked to use a new tool without being told what good use looks like, where human judgment still matters, or how errors will be handled.
The report shows how much manager communication changes the experience. Among workers whose leaders explain how AI will change their role, 92% said they are comfortable with AI-related change. Where that communication is absent, the share drops to 15%.
That helps explain why Fortune’s independent coverage framed the findings as a leadership test, highlighting both the survey’s scale and the low empathy figure. If employees do not know how AI affects their role, then even a technically successful deployment can feel like a moving target: more output expected, less clarity provided.
This is where the study’s retention findings become practical, not cultural. Adecco says work-life balance became the leading reason respondents consider leaving an employer, ahead of salary, with schedule flexibility and remote-work opportunities next. If AI tools are introduced mainly as a speed lever, managers may convert saved minutes into higher output expectations instead of higher-value work. That can raise work intensity even when the software performs as intended.
For employers, that turns AI from a software procurement story into a labor-cost and retention story.
What leaders should measure before calling AI productive
The report is most useful when read as an operating scorecard rather than a referendum on AI itself. Companies scaling agents should measure more than licenses, active users, and automated tasks. Those are rollout metrics, not value metrics.
A more credible scorecard starts with job design. Before expanding an agent, leaders should define the task boundary and the decision boundary: what the system is allowed to do, what remains with a person, and what must be reviewed before action is taken. They should inventory the data and permissions an agent can access, identify failure cases as well as normal cases, and assign a human owner who is accountable for review and recovery when the tool is wrong.
They also need worker involvement early, not after deployment. Adecco reports that 96% of workers whose employers involved them in redesigning work said they were willing to adapt to changing environments, compared with 29% of workers who were not involved. That does not mean participation guarantees success. It does suggest that employee voice is not a soft add-on; it is a way to surface hidden dependencies, judgment calls, and practical obstacles that software teams and senior managers may not see.
Then comes measurement. If leaders want to claim productivity, they should track quality, rework, cycle time, workload, psychological safety, and retention together. A rollout that reduces cycle time while increasing rework or burnout is not yet productive in any durable sense. A rollout that drives heavy use but leaves workers unsure how their role is changing is not mature, no matter how impressive the dashboard looks.
Because Adecco has a commercial interest in training and workforce transformation, its people-centered framing aligns with its business. Even so, the warning is a sensible one. The survey does not show that every employer is making the same mistake, and its global averages may hide big differences by country, industry, occupation, and the kind of “agent” being used. What it does show is where AI adoption can break operationally: workflow fit, skills, manager communication, and employee trust.
That is the real test now facing employers. The next phase of enterprise AI will not be decided by how many workers can access an agent. It will be decided by whether companies treat training, redesign, and management support as part of the product cost of deployment—or as optional overhead to worry about later.




By
By
By
By
By








