Jeff Bezos recently declared that AI is ushering in a “golden age” of the workplace. GoTo’s Pulse of Work 2026 report suggests the reality on the ground is rather more complicated.
The research, which surveyed workers across industries on their AI habits and attitudes, paints a picture of adoption outpacing accountability.
Seven in ten employees now admit to misusing AI for high-stakes tasks, up from 54% just a year ago. Half say they rely on it too much. Two-thirds report that reviewing poor AI outputs creates more work than reviewing content written by a colleague.
For customer experience leaders, the findings should not be ignored.
The “high-stakes tasks” employees are handing to AI include legal and compliance work, decisions involving sensitive customer data, and actions that directly affect people’s lives.
The AI governance gap runs straight through the front line of customer service.
To understand what’s driving the trend, CX Today spoke with David McGlennon, Staff Product Manager at GoTo. He said:
“The trajectory does worry me. It’s telling us that AI is becoming more and more powerful, and businesses are really keen to adopt it, given the efficiencies it produces.
“But they’re not necessarily creating the governance that needs to be set around those tools in that rush.”
That governance gap has several dimensions. McGlennon pointed to the following issues around data access and permissions:
- Whether AI tools are scoped to read-only or can update systems
- Whether they’re handling personally identifiable information
- Whether employees even know how to prompt them effectively
“There’s quite a few things around governance that companies are likely not covering,” he said, “because maybe they don’t have the expertise, but they can obviously see that these tools are so good at producing outputs.”
The Shadow AI Problem
Nowhere is the AI governance gap more acute than with the tools employees are actually choosing.
A recent Gartner study found workers are three times more likely to use third-party generative AI than tools provided by their employer. GoTo’s own research tells a similar story; it shows that 84% of employees say companies aren’t doing enough on responsible AI use, yet only 48% of IT leaders agree.
McGlennon’s explanation for that disconnect is telling. IT leaders, he argues, are operating at a macro level: a policy exists, so in their view, they’re covered. Employees, meanwhile, are working day-to-day with whatever fills the gaps their employer hasn’t addressed.
“Some of the employees may be using tools that are external to what the business has offered them,” McGlennon said. “This isn’t necessarily something that the business has delivered.”
For CX teams, that’s a tangible risk. Customer service agents handling sensitive customer data, drafting complaint responses, or managing escalations may be feeding that information into consumer-grade tools with no enterprise governance, no data retention controls, and no audit trail.
Approval Fatigue and the Agreeable AI
One of the more interesting themes from the conversation was what McGlennon calls “approval fatigue.”
Embedding review steps into AI workflows is designed to keep humans in the loop, but when agents are processing volume under pressure, those checkpoints quickly become rubber-stamping exercises, as he explains:
“You’re approving everything as you go because you’re like, ‘yep, that’s fine, go do it,’ to the point where you’re not even validating it.”
That problem is made worse by one of AI’s less-discussed characteristics: it tends to agree with you.
“AI can be incredibly agreeable,” McGlennon noted. “You need to sort of ask it to be devil’s advocate. Can you second-guess yourself? Can you tell me why my idea doesn’t work?” For CX leaders building AI-assisted service workflows, particularly around sensitive customer interactions, that sycophantic quality is worth designing around explicitly.
A Quiet Concern About the Future Workforce
The report’s Gen Z findings add a longer-term thread to the conversation, with 46% of Gen Z workers admitting that overusing AI is making them less intelligent – the highest share of any generation.
For an industry that depends on human judgment, empathy, and the ability to de-escalate difficult customer situations, that could prove to be a significant issue.
In discussing this point, McGlennon explains that he sees the future of work being defined less by whether employees use AI and more by how well they use it, as he explains:
“The future will be about who’s capable of adopting AI in the right manner. There’s an awful lot of critical thinking that needs to be applied to the output.”
What CX and Business Leaders Should Do Next
McGlennon’s advice to leaders sitting with this data is to start finding out what AI tools your people are actually using, and do it without judgment.
“People are almost certainly going to be using them, and the stats prove that,” he said.
“If it’s an external tool, there is no governance. We need to recognize what that is and be open to understanding that there are tools being used that weren’t on the list of procurement.”
From there, the steps are clear enough:
- Bring those tools into a formal roadmap
- Build in the relevant security restrictions
- Invest in training that covers not just what AI can do but where it falls short
“As an employee, I think they would rather use a business tool versus their own paid subscription.”
For CX leaders, the window to get ahead of this is still open. But that 70% figure, and the direction it’s heading, suggest it won’t stay that way for long.
