Overcoming Employee Resistance to AI

Akhil Verghese, founding leader of Krazimo, explains how companies can help their teams overcome employee resistance to AI. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.

Are your employees resisting the use of AI in your business’s workflows? If so, then you aren’t alone. Many business leaders are encountering difficulties in this area. In my experience helping companies of all kinds implement AI, I’ve found a few tried-and-true techniques to overcome this resistance. Here’s how to turn your skeptics into AI adopters.

The Common Sources of Employee Resistance to AI

Overcoming employee resistance requires understanding why it emerges in the first place. There are three main sources of reluctance to incorporate AI.

First, there’s a natural human impulse to resist anything new. This “if it ain’t broke, then don’t fix it” mindset can be a significant barrier to AI adoption and greater efficiency.

Second, resistance can emerge because of poor products or inadequate onboarding. For instance, employees can get annoyed if they find themselves with a tool that doesn’t actually do what they need it to. They can also get frustrated if they find themselves spending a lot of time correcting AI’s mistakes.

Third, the fear of being replaced by AI can promote employee resistance, especially in jobs that are entirely focused on tasks the AI is being trained to take over, such as customer service. This fear applies less to jobs that require more human judgment and discretion.

Here are some approaches to AI implementation that I’ve seen be successful.

Introduce new technology with interactive training sessions

For employees who dislike change, interactive training sessions can help bring them on board. To make these sessions effective, leaders should:

• Identify 2-3 early adopters who are already successfully using AI.
• Ask them to document one real workflow that AI improves.
• Run live, task-based sessions rather than abstract demos.
• Collect feedback immediately after sessions and iterate.

The basic idea is that when people begin to see how AI can make their jobs easier from someone they know and respect, their resistance tends to evaporate naturally. Change not only becomes possible but also desirable. But that’s not all. In my experience running such sessions, new ideas and even new products can emerge from these conversations, with the potential to drive future business growth.

Ensure the right products.

All too often, business leaders add AI to areas without a thorough evaluation or a clear definition of success. This approach sets employees up for a frustrating experience. Before deploying any AI solution, the manager should know:

• What their current performance is (e.g., 1.3 percent of outbound sales calls close).
• How much it costs (e.g., $1,000 per close in net costs).
• What an acceptable drop in performance would be before cost saving isn’t valuable (e.g., if each close results in $10,000 in profit, then the 1.3 percent rate can drop only so far, even if the cost is zero, before it’s still a net negative).
• What success would look like. For example, a 0.9 percent close rate with $100 per close means you can now make 10x as many calls at the same cost. However, if you’re already maxing out your leads, the reduced cost matters less, and almost no reduction in close rate is acceptable.

Only once business leaders understand what they want to achieve with agentic AI can they find the business solutions that align with their aims. Otherwise, they may fall for a slick sales pitch that doesn’t actually give them the capabilities they need. For instance, in the new “agent washing” trend, many companies market themselves as AI when their products aren’t customized to your specific requirements, don’t integrate with your tools, don’t have adequate guardrails, and are essentially just chatbots.

For this reason, it’s necessary to listen to the employees and managers who are actually using the new tools. In addition, collect data that enables you to assess the effectiveness of the AI resources. If the person most often using the LLM is closing three times as many deals, that’s a clear sign the AI is effective. If usage isn’t resulting in measurable improvements, then evaluate if you’re using the right tools.

Spend time on phased launches.

Incorporating AI into workflows takes time. Most useful systems get up to speed by making mistakes, receiving feedback, and adjusting as necessary, which means they cannot be powered up and let loose without oversight. Instead, they should be gradually implemented over time in a phased process.

Begin by having the machine attempt to complete activities in parallel with human staff. During this stage, a manager should compare the two sets of work to see how the AI’s efforts stack up. Meanwhile, ensure strong guardrails to prevent cost overruns. For instance, the AI should not be allowed to keep retrying tasks it is struggling with; instead, it should stop after a certain number of attempts and ask human staff for guidance.

When the manager is confident that the AI can reproduce the work of a competent employee and won’t cause any harm, the machine can be entrusted with those aspects of the work. As it proves itself, the AI can be given more and more responsibilities, much like a junior-level employee.

Even when AI systems prove themselves to the point that they can function largely autonomously, their actions need to be audited and overseen. AIs should always be required to ask for permission before doing anything that would be expensive or could have legal consequences.

In addition, AI systems need to be maintained like any other piece of equipment. Policy changes over time, prompts get adjusted, and data changes. That’s why it’s crucial to monitor these systems for drift and mitigate against it.

Lastly, look into which teams could be reskilled if their positions were entirely task-based. I’ve seen customer service representatives transition into sales roles, benefiting themselves and the company.

Implement AI the right way.

Companies that fail to learn how to use AI in their day-to-day work are almost certain to be left behind in the next five years. Overcoming employee resistance today will be crucial to unlocking its massive potential tomorrow. It pays to slow down and implement AI systems the right way.