While 88 per cent of surveyed employees say AI has benefited them, 41 per cent believe that leaning on AI too heavily will damage their career in the long run. Image: techfinitive.com.
Dan Schawbel says an overall reliance on artificial intelligence is producing a human workforce that is under-trained, under-guided and prone to using AI inappropriately.
Here is what a productive workday looks like in 2026: An employee opens their laptop, lets AI draft their emails, summarise their meetings, generate a first cut of their report, and flag the data points worth noting.
They review the outputs, adjust, and hit Send. They’ve saved more than two hours. By every measurable standard, they’ve had a great day.
Now ask that same employee how they feel about their own abilities. Ask whether they trust their judgment more or less than they did two years ago. Ask whether they could do their job without the machine.
The answers, revealed in new research by GoTo Technologies, conducted in partnership with my company, Workplace Intelligence, should raise alarm bells for business leaders.
Some 50 per cent of workers say they rely on AI too much, with 39 per cent believing that dependence is actively making them less intelligent.
We are in the middle of an AI confidence crisis that the productivity numbers have been obscuring.
Nearly three in 10 employees feel AI is doing their job better than they can. A similar share says they’ve started trusting AI’s judgment over their own.
While 88 per cent say AI has benefited them, 41 per cent believe that leaning on AI too heavily will damage their career in the long run.
That tension is not a reason to slow AI adoption. It is a reason to take organisational responsibility for how that adoption happens.
The confidence gap doesn’t emerge in a vacuum. It is, in large part, a predictable outcome of how most organisations have rolled out AI — too fast, too broadly, and without enough structural support.
That pressure to use AI, applied to a workforce that is simultaneously under-trained and under-guided, has resulted in 70 per cent of employees admitting to using AI for sensitive or high-stakes work.
This includes legal and compliance tasks, decisions requiring emotional intelligence, strategic choices with real consequences, or situations involving confidential information.
The results land on everyone’s desk with 43 per cent of employees having submitted AI-generated content they suspected was low quality or factually wrong.
As a result, 77 per cent say reviewing AI-generated work takes longer than reviewing work done by a person, with sorting through other people’s AI output creating additional work.
The efficiency gains from AI are genuine, but so is the new category of inefficiency that sloppy AI use is generating.
None of this is primarily a workforce problem. It is an organisational design problem.
Employees aren’t cutting corners because they don’t care about quality. They’re operating in environments where the expectation to produce with AI is high, and the guidance on how to do it responsibly is low.
One of the most consequential findings in the GoTo research isn’t about employee behaviour at all. It’s about the AI perception gap between workers and IT leaders.
While 84 per cent of employees say their organisation could do more to encourage responsible AI use, only 48 per cent of IT leaders agree.
So, while employees are experiencing a daily reality of pressure, under-training and uncertain guardrails, many of their IT leaders and teams believe things are largely fine.
The research doesn’t just diagnose the problem. It points to what responsible AI adoption looks like in practice.
The skills employees themselves say matter most in an AI-driven workplace are checking AI outputs for accuracy and bias, knowing when to trust AI and when to override it, and applying human judgment alongside machine output.
These are collaboration skills: the competencies that let a person work with AI rather than simply defer to it.
Employees also identified the practical tools that would help most: AI support for data entry and email management, communication, administrative tasks, content creation, and client support.
These are areas where AI can absorb volume and repetition, freeing workers to apply the judgment, creativity and emotional intelligence that remain distinctly human.
The goal isn’t AI instead of people. It’s AI handling the load so people can do the work that requires uniquely human skills.
The productivity story about AI is true. Workers are saving real time, completing more, and reporting genuine benefit from the tools now woven into their days. That part of the narrative doesn’t need defending.
What needs scrutiny is the assumption that productivity gains are evidence that everything else is working. They’re not.
Below the surface of a productive workforce is one that is increasingly uncertain, quietly overwhelmed by its own outputs, and missing the training and policy infrastructure to use AI with the skill and judgment it demands.
The organisations that will look back on this period as a competitive advantage are not the ones that deployed AI fastest. They’re the ones that treated deployment as the beginning of a responsibility to train workers, set clear boundaries, model thoughtful use at the leadership level, and build the kind of human-AI partnership that makes people more capable, not less.
That work doesn’t happen automatically. It requires someone to own it, but right now not enough organisations are taking that responsibility.
Dan Schawbel is a bestselling author and managing partner of Workplace Intelligence, a research and advisory firm helping HR adapt to trends, drive performance and prepare for the future. This article is part of his Workplace Intelligence Weekly series.
