Employees Will Work With AI Agents. They Won't Work For Them Yet

Ask most vendors and the pitch is the same: AI agents are ready to take real work off your plate. Ask the people who'd actually be handing that work over, and the answer is more careful. New global research from Workday, "AI Agents Are Here — But Don't Call Them Boss," puts a number on exactly where that caution sits: 75% of employees say they're comfortable teaming up with an AI agent. Only 30% say they're comfortable being managed by one.

That's not a rejection of AI agents. It's a very specific line being drawn about what role they're allowed to play.

The trust gap, in the workers' own words

Workday's research, fielded by Hanover Research across a global survey of 2,950 full-time decision-makers and software implementation leaders in North America, APAC, and EMEA, breaks the trust question down further than the headline stat suggests:

• Co-pilot, not commander. Three out of four employees are highly comfortable with AI agents working alongside them or recommending new skills. Comfort drops sharply — to 30% — the moment the agent is positioned as managing them.
• Visibility matters as much as authority. Only 24% of employees are comfortable with an AI agent operating in the background without their knowledge, even doing the same task they'd otherwise be fine with if they could see it happening.
• Task type changes everything. Trust is highest for IT support and skills development, and lowest for hiring, finance, and legal matters — the categories where a wrong call has the biggest downstream consequences for a real person.
• Exposure builds trust fast. Only 36% of employees who are still "exploring" AI agents trust their organization to use them responsibly. Among employees further along in actually using them, that jumps to 95%. Direct experience, not marketing, is what moves the number.

Kathy Pham, Workday's VP of AI, frames the takeaway plainly in the release: "Building trust means being intentional in how AI is used and keeping people at the center of every decision." The data backs that up — the boundary employees want isn't "less AI," it's "AI I can see and understand, doing work I've agreed to hand over."

The productivity trade-off nobody's hiding from

The Workday data doesn't paint AI agents as a hard sell, either. Nearly 90% of employees believe AI agents will help them get more done — that's not a skeptical workforce. But the same survey shows people are clear-eyed about the cost: 48% worry the productivity gains will come with increased pressure, 48% worry about a decline in critical thinking, and 36% worry about less human interaction at work. Meanwhile, 82% of organizations report they're already expanding their use of agents regardless — the adoption curve isn't waiting for every concern to be resolved first.

Finance is a useful case study in how this plays out when trust is actually high. With the industry facing a well-documented shortage of CPAs and finance professionals, 76% of finance workers in the survey believe AI agents will help close that gap, and only 12% are worried about losing their jobs to one. The top uses they report for agents in finance are forecasting and budgeting (32%), financial reporting (32%), and fraud detection (30%) — concrete, bounded tasks with a clear owner still reviewing the output, not open-ended autonomy.

What "delegation" should actually mean right now

Put together, the Workday findings sketch a fairly specific and workable version of delegation, not a wholesale handoff:

1. Keep the human in the loop on anything with real consequences. Hiring, finance decisions, and legal matters are exactly the areas where trust is lowest — start agents on lower-stakes, well-bounded tasks and expand from there, not the reverse.
2. Make agent activity visible, not invisible. The 24% comfort figure for background agent activity is the clearest signal in the data: employees aren't asking for AI to disappear into the workflow unannounced, they're asking to know when it's running and on what.
3. Let usage build the trust, don't try to argue people into it. The jump from 36% to 95% trust with direct exposure suggests the fastest path to adoption is hands-on, supervised use — not a policy memo explaining why agents are safe.

What this means for teams rolling out agents

The organizations that get the most out of AI agents in 2026 probably won't be the ones with the most aggressive rollout timeline — they'll be the ones that matched the agent's role to where employees already have real trust, and let that trust expand naturally as people got hands-on experience. Workday's own data makes the ceiling clear: the demand for AI agents as teammates is already there at 75%. The work left is proving, task by task, that a teammate is really all they are.