Can force adoption solve AI resistance?

Short-term AI mandates help long-term adoption, only if the results are visibly rewarding

Featured faculty: Cao Xinyu
Written by Sally Ho

Across industries, companies are integrating artificial intelligence (AI) tools to support decision-making and day-to-day operations. It wouldn’t be wrong to assume that once employees understand the benefits, they will automatically embrace AI, but this is often not the case.

Even though AI technology delivers positive results, firms are still struggling to persuade employees to embrace it. A 2026 global survey from the digital adoption platform WalkMe finds that more than half of white‑collar employees abandon their AI tools and revert to manual work.

According to Cao Xinyu, Vice-Chancellor Associate Professor of Marketing at the Chinese University of Hong Kong (CUHK) Business School, algorithm aversion, or a human tendency to distrust computer algorithms, even when they perform better, has been well documented.

Sceptics view machines as inferior to humans, leading them to use AI only for mundane and laborious tasks. Employees may also worry about accountability for algorithmic errors or feel their professional abilities are undervalued.

Such concerns are reasonable, but complete resistance can prevent employees from realising the true benefits of the technology. It can also undermine AI investments, especially when the management has spent a fortune on tools that are used intermittently.

Companies often rely on training, which sometimes includes workshops on the practical application of AI, and encourage employees to use AI tools to improve productivity, but such methods aren’t sufficient. At this point, you may joke that forcing employees to use AI could work, but Professor Cao accidentally confirms it.

“We did not anticipate that temporary mandatory use would lead to sustained AI adoption. Our initial goal was simply to obtain a fair performance evaluation of the AI tool, but we then observed an interesting behavioural change among our research participants, so we explored its underlying mechanism.”

How temporary AI use leads to lasting adoption

In a study titled How forced intervention facilitates AI adoption, Professor Cao and her co-authors, Hu Chenshan of the University of Colorado Boulder, Sun Jiankun of Imperial College London, and Dennis Zhang of Washington University in St. Louis, collaborate with a large online education company in China.

The education platform introduced an AI tool to help sales staff suggest trial class teachers for prospective students. Despite its simplicity and convenience, this tool was underutilised. On average, sales staff use the AI tool only for 20 per cent of the prospective students. Furthermore, they tend to use the AI tool for low-quality leads who are less likely to convert.

“By transparently showing improvements, firms can help correct workers’ biased beliefs about AI and reduce resistance to adoption.”

Professor Cao Xinyu

To investigate the root cause, Professor Cao and the team evaluated the AI tool’s performance by dividing 171 sales employees into three groups. For three weeks, one group was asked to use the tool, another was prevented from using it, and the third group was free to use AI or not.

The AI tool is found to perform comparably to humans in terms of conversion rates, while significantly reducing manual effort and increasing efficiency. Those subject to a short-term mandate are more likely to continue using the AI tool, even after the experiment ends. Some may argue that repeated use makes employees become familiar and eventually form a new habit, but the data suggest more than that.

“If it’s a habit, everyone who was forced to use the AI tool for a while would keep using it in a similar way afterwards, but that’s not what we found. Employees who experienced a larger increase in conversion rate during the mandate-AI-use period tend to use AI more and use AI more on high-quality leads after the experiment,” Professor Cao adds. “Habit alone wouldn’t explain such varied behaviour linked to individual results.”

First-hand experience changes perception

Direct experience allows employees to reassess and change their biased beliefs about AI accordingly. People gain knowledge and skills through real-world practice, which can be more influential than traditional methods such as reading or listening.

The study also offers a clear lesson for company managers. Sometimes, obstacles may not stem from AI tools’ capabilities, but rather from how employees perceive the new tools. When scepticism is high, voluntary uptake alone may not be enough. A fleeting period of structured use allows users to evaluate the system on their own terms.

The caveat, Professor Cao warns, is that mandatory use must be implemented carefully. If employees feel pressured without a clear context, resistance may increase instead of decline. The main goal is not compliance but enabling an informed experience. She also suggests that firms should complement exposure with tangible results. “By transparently showing improvements, firms can help correct workers’ biased beliefs about AI and reduce resistance to adoption.”

Future challenges in AI adoption

As AI becomes increasingly embedded across industries, Professor Cao believes the findings are relevant to other sectors. “Biased beliefs about AI are quite common across industries, especially during the initial deployment in organisations. Employees who hold biased beliefs towards AI tend to underutilise it, and revising such biases plays an important role in addressing algorithm aversion.”

Beyond AI adoption, businesses anticipate the next phase of industrialisation, Industry 5.0, in which technology shifts from digital automation to a human-centric model. Companies will continue to face questions about how employees interact with algorithms, and for Professor Cao, this means more questions to answer.

“For example, how to best allocate tasks between human workforce and AI, how to design the incentives for employees in using AI, how AI transparency influences humans’ trust, learning, and long-term adoption,” she says. “These directions can help deepen our understanding of both the behavioural and market-level implications of AI adoption.”

For now, the study suggests that organisations may need to focus not only on technical deployment, but also on how employees learn to work with technology. In some cases, a brief period of direct experience may be enough to change how workers think about a technology and whether they choose to use it for good.