Many large companies are investing heavily in artificial intelligence tools in the belief that they will enhance decision-making, spark innovation and help leaders boost employees' productivity. But despite this, studies show that they are failing to achieve the expected benefits, with as many as 80 percent of companies reporting a failure to benefit from the new technology.
But according to new research from Aalto University in Finland the failure of AI tools to deliver on their promise is not so much about the technology and its capabilities, but more a question of the different emotional and behavioural reactions employees develop towards AI and how leaders can manage these reactions.
"Often employees fail to embrace new AI and benefit from it, but we don’t really know why," says Assistant Professor Natalia Vuori, one the authors of the study. "Our limited understanding stems partly from the tendency to study these failings as limitations of the technologies themselves, or from the perspective of users’ cognitive judgments about AI performance, she adds.
"What we learned is that success is not so much about technology and its capabilities, but about the different emotional and behavioural reactions employees develop towards AI - and how leaders can manage these reactions."
Her research team followed a consulting company of 600 employees for over a year as it attempted to develop and implement the use of a new artificial intelligence tool. The tool was supposed to collect employees’ digital footprints and map their skills and abilities, ultimately building a capabilities map of the company.
The results were supposed to streamline the team selection process for consulting projects, and the whole experiment was, in fact, a pilot for AI software they hoped to offer their own customers. But after almost two years, the company buried the experiment — and the proposed product.
So what happened?
It turns out, although some staff believed that the tool performed well and was very valuable, they were not comfortable with AI following their calendar notes, internal communications and daily dealings. As a result, employees either stopped providing information altogether, or they started manipulating the system by feeding it information they thought would benefit their career path. This led to the AI becoming increasingly inaccurate in its output, feeding a vicious cycle as users started losing faith in its abilities.
"Leaders couldn’t understand why the AI usage was declining. They were taking a lot of action to promote the tools and so on, trying to explain how they use the data, but it didn’t help," says Vuori, who believes this case study reflects a common pattern when it comes to AI uptake, and tech adoption generally.
The same team is now collecting data on the use of Microsoft’s widely used Copilot AI software, which is so far yielding similar findings.
So what should leaders do?
Researchers found that people fell into the same four groups in terms of their reaction to the new technology. Distinguishing between cognitive trust; whether a person believes the technology performs well, and emotional trust; their feelings towards the system, the groups were: full trust, full distrust, uncomfortable trust and blind trust.
People in the first group had high trust both on the cognitive and emotional level, whereas people in the second group scored low on both. Uncomfortable trust signified high cognitive trust but low emotional trust, and vice versa for blind trust.
The less people trusted the tool emotionally, the more they restricted, withdrew or manipulated their digital footprint, and it was particularly notable that this held true even if they had cognitive trust in the technology.
The findings suggest that successful AI adoption should also involve more personalised training for employees as well as more carefully-considered leadership styles.
"AI adoption isn’t just a technological challenge, it’s a leadership one. Success hinges on understanding trust and addressing emotions, and making employees feel excited about using and experimenting with AI," says Vuori. "Without this human-centered approach, and strategies that are tailored to address the needs of each group, even the smartest AI will fail to deliver on its potential."