Skip to Content
Skip to Footer


Why AI Adoption Starts with Psychology

Artificial intelligence (AI) has the potential to dramatically transform how we work today. But studies show that there’s often a significant gap between organizations’ ambitions and their success in implementing AI technology. In fact, up to 53% of AI projects stall in the pilot phase or the early stages of adoption. [Click to Tweet]

What’s going wrong here? Salesforce has found that for many companies, the biggest hurdle in adopting AI isn’t the technology — it’s the people. We sat down with Jess Shutt, Lead User Researcher on Salesforce’s Einstein and AI team, to get a closer look at the issue.

What the research reveals about AI adoption challenges

Shutt’s team recently conducted research with companies of varying sizes and industries, so she has a ringside view on what works when it comes to AI adoption and what doesn’t. A major insight? The key to success starts with understanding human psychology and the way people typically respond to change, then applying effective change management tools.

“Let’s say a team builds an AI model they feel comfortable with. Now they want to move forward, but they don’t know how to progress from having the model in a product and ready to go, to getting a team to adopt it,” Shutt says.

“For example, they may not realize how to find allies who can help them communicate the value of the new AI process and allay any potential job-loss fears. That’s a critical step and without it, employee resistance and a lack of management support can result in the project stalling.”

Another common assumption companies make is that because the AI tool they’re looking to implement is cutting-edge, adoption will simply happen without the need to launch any specific change management initiative.

“A lot of people just assume that if the technology is good enough, teams will just absorb it. It’s almost like attributing AI adoption to pure luck — that there’s nothing people need to do to help the process along,” Shutt explains.

It turns out that the opposite is true.

Understanding the psychology behind it: Status quo bias and psychological inertia

Psychologists have found that it’s natural for people to actively resist adopting any new process. There are two key factors at play here: status quo bias and psychological inertia.

As Shutt explains, status quo bias is what happens when people express a preference for the way things are being done currently.

“Studies show that even if a new process is better than an existing one, most people will still prefer the latter,” she says. “For example, people will generally continue with a more long-winded process — like routinely sending someone an email to request information — rather than checking a box on a form that automates that request. There has to be quite significant and obvious value in any change before people adopt it.”

Psychological inertia is the related idea that just as it’s hard to change the course of a train once it has developed momentum in a specific direction, it’s equally difficult to get people truly invested in change once their ideas, processes, and habits have solidified.

“The good news is that if people and teams are expressing inertia toward making a significant change in the way they work, we can use scientific processes to ensure a successful redirect — just as we would if we were redirecting a train,” says Shutt.

AI is new but the old adage of “put people first” still applies

In Shutt’s view, human vision, guidance, and input ultimately play a big part in any AI project’s success. For any rollout to be effective, it’s vital to prioritize people, not technology.

“AI is changing how we think about day-to-day work,” she says. “But we shouldn’t forget the lessons we’ve learned in previous periods of great technical disruption about the way humans react to change. Success comes from arming people with change management best practices to help their teams and business adapt.”

Effective starting points for any change team to consider include:

  • Understanding what your target audience cares about most: What drives the corporate culture in the team where you’ll be implementing AI? If you’re not thinking about what’s most important to people, there’s a risk your change initiative may miss the mark.

    “Recently, I worked with a call center team where the reps were highly motivated to adopt an automation solution because they believed it would help them meet their goals of answering customer calls more efficiently and accurately,” Shutt recalls. “In short, they already had a great relationship of trust with data. That meant it was easy for the change team to refer to data and numbers during rollout to explain the type of situations where the AI would work well and where it wouldn’t.”

    However, other teams may not have that same level of trust in data. “Such an approach wouldn’t work well with teams where the culture is more focused on relationship building,” Shutt continues. “In such cases, a successful AI rollout would need to focus on showing people that if AI can automate a number of day-to-day tasks, it would free them up to focus on those important client-building relationships.”
  • Forming powerful coalitions: Identify the people within your company who can act as allies. “One way is to pick a single team to trial the AI process — the bigger and more high-profile, the better,” says Shutt. “Their success in implementing the process will act as a mini case study as you roll out the AI across the rest of the enterprise.”

    Another effective approach is to identify a single ally from every relevant team — whether that’s sales or service, commerce or marketing — across the enterprise. “That way, a global company might end up with a 30-person group of pilot users from different roles, companies and territories across the world,” Shutt says. “If every team has a strong ally embedded within it, it can be effective in proving that the AI process is applicable to every part of the company.”

  • Creating short-term targets: Isolate and communicate a series of achievable smaller targets as you implement your AI solution, not just one major long-term goal. For example, start by trialing a limited version — the equivalent of a minimum viable product (MVP) — to see how your target audience will react. Based on the feedback you gather, you can create incremental improvements to make the basic functionality and the user experience better. That way, you’re making the adoption process as inclusive as possible, which can further motivate the entire company to adopt the solution in questions.

As Shutt puts it, “AI adoption isn’t a hurricane. It’s not beyond our control.”

“By falling into the trap of thinking there’s nothing we can do to help adoption happen because AI is new, we diminish our influence over the situation,” she says. “In fact, there’s no need to reinvent the wheel. The tried-and-tested change management principles that have worked well for companies over decades all still apply.”