From data to insight to action: The very human challenges of AI transformation

A few short years ago, the idea of collecting a million data points per day during any process was unfathomable to most organizations. Now, with the advent of powerful acquisition methods and affordable storage options, we’re awash in data. The challenge is sifting insights from this deluge, and then converting them into actions that transform processes and organizations.

That’s where AI can help. No matter the industry, AI’s unprecedented ability to analyze and identify patterns in data promises to radically change how organizations operate, such as making sales calls more productive, reducing waste in factories and saving lives in high-risk industries. But to accomplish true AI transformation, we need to understand humans more than we need to understand the technology.

As cognitive scientists, we’ve noticed that AI transformation comes in three stages: collect data, find insights and take action. The latter two stages require a deep understanding of what drives human behavior: The fears, motivations, biases, cognitive capacity limitations and other brain processes that cause people to act a certain way. AI can identify patterns in data, but to derive insights from the patterns and then design effective organizational change initiatives, understanding humans is imperative.

Using AI to save lives

Let’s examine the three-step process of AI transformation with a real-world example. Dr. Teodor Grantcharov, professor of surgery at Stanford University, wanted to use AI as a tool to analyze, and hopefully decrease, surgical errors in the operating room. Although estimates vary widely, studies suggest that between 44,000 and 250,000 patients die in the U.S. each year due to medical error. About one-fourth of those deaths occur because of preventable mistakes in the operating room (OR), studies have estimated.

For 20 years, Grantcharov has been developing an “operating room black box” that analyzes everything that happens during a surgical procedure. He drew inspiration from the flight data, or “black box” recorders used on airplanes. Since 1957, when the U.S. Civil Aeronautics Board mandated flight data recorders on all passenger aircraft, the instruments have helped illuminate the causes of accidents and disasters. Black box recorders have saved countless lives through changes to pilot training, airline equipment and regulatory standards. 

The OR black box was developed with a similar purpose in mind: Identifying and then taking actions to mitigate preventable errors. In recent years, improvements in AI have allowed Grantcharov’s team to overcome their former bottleneck of data analysis. The insights they gained significantly enhanced individual and team performance and increased compliance with standard operating procedures. These changes reduced morbidity, mortality and costs in operating rooms that used the black box, Grantcharov says.

Step 1: Collecting data

The first step in AI transformation is collecting data, which today is the easiest step. So far, Grantcharov has placed the platform in around 20 operating rooms across the U.S. Through a variety of sensors, the OR black box captured up to 1 million data points per day per site. These included audio-visual data of surgical procedures, electronic health records and input from surgical devices. The data also included biometric readings from the surgical team, such as their heart rate variability as a reflection of stress levels, and brain activity measured by wireless EEGs. 

The data contained a wealth of information, but according to Grantcharov: “Data is useless if we can’t turn it into information that clinicians can use to change their behavior.” 

Step 2: Finding insights

Identifying patterns in data is where AI is particularly helpful. “It’s impossible for the human brain to constantly monitor all these data points and look for patterns and hidden associations,” Grantcharov notes. “That’s where modern AI methodologies can really empower us to turn data into insights into action.”

But here’s where it’s also important to understand humans. AI can correlate OR accidents with certain events, but without a working hypothesis, it’s all just noise. For example, Grantcharov’s team hypothesized that stress could affect a surgeon’s performance by impacting their cognitive processing and decision making. So they designed the experiment to collect physiological data from the surgeons, and AI was able to correlate these data with OR accidents. The finding: Stressed-out surgeons had a 66% higher chance of making an error. 

Grantcharov also noticed events like a door opening, a phone ringing or somebody talking about last night’s football game — in other words, distractions — were the root cause of the most catastrophic errors. Finding this insight required an understanding of the brain’s finite cognitive capacity.

Deriving other insights required an understanding of team dynamics. The researchers observed teams that communicated poorly and lacked psychological safety — the belief that they could speak up and raise concerns when necessary — had worse outcomes regardless of the surgeon’s level of technical skill. “One of the most dangerous operating rooms is a silent one, where nobody is speaking up or communicating,” says Grantcharov.

Although one might assume that the surgeon’s skill is the most important determinant of success, the non-technical attributes of a surgical team, such as how they collaborated, or whether they felt safe to voice concerns, most strongly affected patient outcomes. “It all comes down to culture,” says Grantcharov. 

Step 3: Taking action

Once AI helped reveal the biggest sources of OR errors, hospitals and surgical centers could, at least in theory, begin introducing new procedures to prevent mistakes. But first, they had to understand how behavior change happens. Successfully changing an entire organization’s culture requires the establishment of priorities, habits and systems,

Priorities are the tasks or activities deemed most important to an organization, and it’s essential to communicate these priorities so everyone knows where to focus their time and attention. In this case, the priority is clear: Improving patient outcomes by avoiding preventable OR mistakes. 

Habits are behaviors that are performed automatically with little conscious thought. For example, speaking up with concerns, instead of remaining silent, can become a habit with training and practice. 

Finally, systems are procedures or principles put into place that make the desired behavior the easiest to do. For example, to reduce distractions and preserve cognitive capacity, hospitals could institute a new rule that restricts non-relevant discussions during critical steps of a surgical procedure. 

Along with priorities, habits and systems, AI transformation requires everyone in the organization to embrace a growth mindset — the belief that failures are opportunities to get better, rather than threats to one’s standing or status. Grantcharov recalls that at first, many surgical teams were wary of the OR black box, worrying that it would make them look bad or leave them vulnerable to litigation. But gradually, their attitudes changed. 

“Once we realize that we can’t improve without objective measures of our performance, it really opens the world of growth mindset and continuous improvement,” he says. Hospitals that welcomed this transition have realized tremendous gains, not only in quality and safety, but also in efficiency and productivity, he claims.

Beyond the OR

Not every industry has as much at stake in terms of human life as the healthcare industry. Yet no matter the sector, AI can analyze data and lead us to valuable insights that drive action, from improving a specific process to changing an entire culture. However, just pointing AI at a data set will reveal little, without a hypothesis worth testing.

For example, in a meeting setting, AI-powered devices could collect audio and visual data (in an anonymized and ethical fashion), and, with the help of human insights, detect patterns that might not be obvious: Are there quiet people in the room who have great ideas, but others constantly talk over them? Is anyone showing signs of excessive anxiety or stress? Are people looking down often in a video call, possibly distracted by devices? 

In this way, AI could help leaders first recognize obstacles that get in the way of productive meetings, then find ways to address them, such as working to increase psychological safety or decrease distractions.

Whether in the operating room or the boardroom, AI can help unlock potential in your organization. But ironically, the more technology plays a central role in our lives, the more we need to understand how humans interact with and process the world.

Dr. David Rock coined the term neuroleadership, and is co-founder and CEO of the NeuroLeadership Institute (NLI).

Laura Cassiday is managing editor of content at the NeuroLeadership Institute.