Shriram Rajagopal is Chief Product Officer at VitalEdge Technologies.gettyAfter launching a purpose-built AI platform for heavy equipment dealers, our team quickly discovered an unexpected issue: It was almost too good.​I met with a senior mechanic who had spent decades learning how to fix every machine in their dealership. When a customer complained that an engine was spewing blue smoke, our AI tool generated the entire repair ticket in minutes. Every detail was right, according to the tech, who was unnerved by its accuracy.​As he said, “Be careful what you wish for.”​But instead of scaling back the AI, we made adjustments grounded in human behavior and psychology.​Fast Adoption Revealed New Opportunities​Even though I’m in a more traditional industry, AI adoption has been faster than even I expected, thanks in part to FOMO. Everyone wants to find more ways to do more with less, get work done faster, and increase the number of employees who can do high-value work.​Our team created an AI platform purpose-built for our industry, trained on high-quality data specifically for heavy equipment dealerships. With 60 years of industry expertise, we knew exactly what our customers were looking for—or at least we thought we did.​Counterintuitive Lessons In AI Expectations And Adoption​AI is a technological solution that also requires a change management strategy, since it often elicits an “emotional, rather than logical, response.” Here are a few specific ways to ensure greater adoption when deploying new AI initiatives.​1. Give People Opportunities to Add Their Judgment and CreativityAI can analyze vast amounts of information much faster than humans, which is incredibly helpful in cases like ours, where technicians need access to hundreds of different equipment manuals.But, as Microsoft stated, humans are “uniquely capable of creativity and judgment,” which isn’t always captured in AI regardless of how comprehensive it is. In conversations with customers after our AI launched, we heard that it was an incredibly helpful tool, but there were still some areas where technicians wanted to rely on their own insights and judgment.​The gap wasn’t that the AI made mistakes. It was that the AI didn’t always have enough ways to learn from what technicians already knew. The technicians weren’t saying “don’t use the AI”—they were just asking to leave room for more of their input, which historically happens person-to-person.​Solutions:• Ask senior people to share their expertise on an ongoing basis to keep improving the AI; consider using voice-to-voice technology—recording calls or in-person conversations and feeding that expertise back into the model.• If a user rejects the AI’s recommendations multiple times, the AI should encourage them to talk with a colleague or supervisor to get their advice.​2. Make Sure Employees Still Have Control Over Decision-MakingPeople want to feel smart and in control. However, researchers have found that AI can lead to anxiety among employees, and half of Americans are “more concerned than excited” about AI in their daily lives, according to a Pew Research Center survey.So we asked a different question: Who actually makes the decision? We decided the technician always signs off, even on repairs the AI got right on its own. That’s not about the AI’s accuracy—it’s about who you trust.Ultimately, we want people, not software, to be accountable, especially in a business that’s built on relationships. Keep people’s feelings and emotions in mind as you build AI solutions, since “a team’s EQ is often as important as the technological aspects.”Solutions:​• Be cautious about letting AI do 100% of the work. After the AI makes recommendations, give the user a list of five things they need to pay attention to, or opportunities to confirm (or correct) what the AI is telling them.• Consider a separate customer-facing version of the AI output if you’re concerned about justifying your team’s value. For example, some of our dealerships offer a streamlined version of the repair ticket rather than the full AI-generated detail.​3. Show Employees That AI Doesn’t Have to Be “All or Nothing”​A technician was out on a service call for a forklift when the client asked them to look at another forklift from a different manufacturer. Because the other forklift wasn’t part of the technician’s inventory, their AI platform didn’t have the parts list and other data to repair it properly. The technician became frustrated going back and forth between multiple software platforms, and ended up not using the AI to fix either forklift.​If an AI platform can’t do everything, some people won’t use it for anything. In our case, once technicians stop using it for a few pieces of equipment, they’re less likely to use it even when they should. People want one tool they can trust for everything, and most AI solutions aren’t there yet—especially those purpose-built for specific industries.​Solutions:• Make sure people understand the value that AI brings to specific problems. If an AI solution is designed to solve problem “X,” that doesn’t mean employees should abandon it just because it can’t solve problem “Y.”• Encourage the use of both AI and non-AI tools side by side. A 700-page repair manual isn’t any less useful today than it was a year ago. We’ve simply developed more helpful tools to supplement them. ​Viewing AI Through The Human Lens​Every AI implementation is a human challenge wrapped in a technology presentation, and requires balancing everyone’s needs, from the CEO all the way down (as PwC noted, “What executives see as reallocating skills, employees experience as a threat to their jobs and expertise.”).The technicians who use our platform every day aren’t asking for a smarter computer to do the work for them. They want tools that support their job as experts, just as they might grab a specific wrench or screwdriver from their workbench. That’s the real challenge for anyone deploying AI at scale—creating a solution that’s better and faster than humans, while still making it something that people want to use.Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?