A Single Memo That Reduced Infection Rates from 11% to 0%: The Miracle of a Checklist [AI Mistake Note]

Hundreds of Millions Saved with a Simple Checklist
The Power of Rules: Guarantees Minimum Outcomes
But There Are Limits to Innovation...
Rules Become Obstacles to AI Adoption

Editor's NoteExamining failures is the shortcut to success. 'AI Mistake Note' explores cases of failure involving AI-related products, services, companies, and individuals.

A person sitting in front of a laptop checking off boxes on a checklist. Photo by Getty Images Bank

"Are there really 40,000 people dying from this kind of infectious disease every year just in the United States?"

Peter Pronovost, professor of anesthesiology at Johns Hopkins Hospital, was surprised to learn that 40,000 people die each year from intravenous line infections.

Internal data revealed that 10% of patients in the hospital's intensive care unit suffered from infections. Patients came to the hospital for treatment, only to contract new diseases?a truly ironic situation.

Professor Pronovost sought a solution. And, surprisingly, he succeeded. Did he develop a vaccine on his own? Or invent a groundbreaking medical device? No. The answer was a single sheet of paper?a checklist.

He organized five items into a checklist.

Actions to Reduce the Risk of Infection

1. Wash hands with soap.
2. Clean the skin with disinfectant.
3. Use a sterile drape.
4. Wear a mask, gloves, cap, and gown.
5. Apply a sterile dressing to the procedure site.

These are basic things every medical professional knows. Yet, in the fast-paced environment of daily hospital life, it was not uncommon to forget one or two steps. Professor Pronovost distributed the checklist to the medical staff and asked them to strictly follow it.

Amazingly, after the checklist was implemented, the infection rate dropped from 11% to below 1% over 27 months. During this period, there were only two cases of infection. Thanks to the checklist, the hospital estimated that it prevented 43 infections and 8 deaths, and saved about $2 million.

Example of Boeing 777 checklist. Boeing

There is another remarkable example that demonstrates the power of checklists.

In 1935, the U.S. Army Air Corps sought to introduce the state-of-the-art B-17 bomber. However, a crash occurred. It seemed that the complexity of integrating advanced technologies made the aircraft difficult to operate. The investigation concluded that "the aircraft was too complex and difficult for pilots to handle."

The introduction of the B-17 was in jeopardy. The alternative was a "checklist"?a list of principles pilots had to follow when operating the aircraft. After that, the B-17 flew 1.8 million miles without a single accident, and achieved remarkable results during World War II.

The Power of Rules: Guarantees Minimum Outcomes, Prevents the Worst

A person is marking a checkbox on a checklist with a pen. Pixabay

Companies and organizations create rules. Checklists are one form of these rules. Human memory and attention are not perfect. Especially under stress, it is easy to overlook important things. Checklists compensate for these human weaknesses.

Rules and checklists ▲prevent missing critical steps; ▲help maintain focus even in stressful situations; ▲ensure that everyone works in a consistent manner; ▲and reduce outcome variability between individuals.

In this way, rules and checklists help minimize errors and achieve consistent results. They are especially important in fields where mistakes can have catastrophic consequences, such as aviation, healthcare, and nuclear power plants.

The Limits of Checklists: No Room for Innovation

A person is recording and checking a checklist in a notebook. Pixabay

Despite these advantages, checklists and rules have critical drawbacks.

Every human is different. Situations vary as well. Applying the same rules to every case can be inefficient. Rules often come at the cost of efficiency. While rules reliably prevent the worst-case scenario, they can also become unnecessary constraints.

Rules undermine adaptability to change. Once established, rules are difficult to modify. It takes time to update them to fit new situations.

As a result, they fundamentally block the possibility of better judgment and decision-making. Where rules exist, there is no need for judgment?just follow the rules. And that is often the safest choice. If you followed the manual, you can avoid responsibility even if the outcome is poor.

However, such situations are hardly ideal. This means missing opportunities for better decisions that take into account the uniqueness and uncertainty of each situation.

AI's Core Function?Prediction?Is Hindered by Rules

An image symbolizing humans and AI. Getty Images Bank

So, what is the essence of artificial intelligence?

As discussed in the sixth episode (the executive who said, "Let's make something with that AI too"), it is "prediction."

AI recognizes patterns, analyzes data, and forecasts future situations. Through this, it creates value by enabling better decision-making.

For such AI systems to function effectively, a key prerequisite must be met: once a prediction is made, it must be able to influence actual decision-making. If the best possible prediction cannot be acted upon because it conflicts with the manual, that prediction is useless. In other words, in an environment where everything is fixed by rules, the value of AI's predictions is limited.

One of the biggest obstacles to AI innovation is existing "rules." This is the critical point where companies seeking to adopt AI face challenges and often become frustrated.

The COVID-19 pandemic is a clear example of the tension between rules and prediction.

When the pandemic broke out, countries around the world implemented strict rules like "social distancing." Movement was restricted, and schools, workplaces, and stores were all shut down. The economy and society came to a complete halt. These rules certainly helped slow the spread of the pandemic, but they also caused enormous economic and social costs.

What if AI-based prediction systems had been used more actively? It would have been possible to predict individual and regional risk levels more accurately and respond in a more differentiated way. The global economy would not have had to shut down all at once?some activities could have continued. Medical resources could have been allocated more efficiently, preventing healthcare system overload. In this way, a better balance between disease control and the economy could have been achieved.

'From Rule-Based to Decision-Based'

People are having a meeting while looking at a tablet PC on the desk. Photo by Getty Images Bank

Checklists and rules have long been the best way to manage uncertainty. They still play an important role in many fields and in our daily lives.

However, in the era of AI, we can manage uncertainty more effectively through prediction, rather than by locking it down with rules. Of course, this does not mean all rules should be abolished. There is no need to gamble. Company policies, contracts, treaties, and various rules have irreplaceable value.

Instead, we need to reaffirm the limitations of rules and redefine their role and scope.

▲Distinguish core rules: Not all rules are equally important. We can identify which rules are truly essential and which can be adjusted.

▲Understand the purpose of rules: Consider what each rule is intended to achieve. If there is a more effective and clearer way to reach the goal, it is okay to bypass the rule. Following rules should never be the goal itself.

▲Gradual transition: Eliminating all rules at once is risky. Start by testing AI decision-making in low-risk areas and gradually expand its scope.

▲Measure and compare outcomes: Objectively compare and evaluate the results of rule-based and AI-based approaches. There is always a possibility that the AI approach could produce worse outcomes.

For companies to truly embrace AI innovation, it is essential to review existing rules and decision-making processes. They must examine whether rules are limiting the value of AI predictions, or whether they are missing opportunities to make better decisions through prediction.

The future is likely to belong not to companies that rely on rules, but to those that predict more often and more accurately. Companies that adapt to this change will reap the greatest rewards from the AI revolution.