The AI That Heals the System

The speed of the technology is not the only measure of the potential of AI relative to the health care field. What was the deepest issue in health care was not a problem that could have been solved with more speed. It is a system that has come apart at the seams. The work that people value and the systems created in service of that work have drifted out of step. AI cannot close that gap. What it can do, used as intended, is help us see that gap, quantify it, and help us begin to bridge it.

That is a different goal from automation. Automation, the easier and less valuable thing, makes an existing process run faster. The harder and more valuable thing is to make a broken system whole again, by getting its various components working together rather than in opposition to one another. When an organization develops a certain way of thinking about AI, rather than merely installing another piece of technology, it begins to do that more valuable work. It begins to stop being a tool that is bolted onto a broken system and begins to be a way that the system is held together.

Disconnection is the Real Wound

It is useful to identify the right wound. The challenge with most health systems is not that a singular component is broken. Rather, it is that the components are disparate. Systems for clinicians and systems for technology steer in different and often conflicting directions. So too are operational and human elements of care. The demand for new systems and the imperative to use systems in a responsible manner are seen as conflicting. Each disconnect is an inconvenience of a minor degree. Cumulatively, they are the cause of the greatest frustration for all stakeholders of the system.

One can observe this phenomenon in discharges, and there is nothing remarkable about that. The physician makes a determination that the patient is ready to be discharged. The nursing staff and the pharmacy both work off disparate lists, and the patient leaves with perhaps conflicting or at a minimum incomplete instructions. Each system in isolation did its job. The problem occurred in the void between them, where no system or tool held responsibility. Pair that with the disjointed systems in a typical organization, and you can appreciate the daily friction every person in it feels, the frustration that seems to have no single cause.

Coherence is the Goal

The term I have chosen to represent the overarching theme of my analysis is coherence. A coherent system is one where all elements work in concert. A coherent system is one where the data supports the decision, and the tool fits the purpose. Coherence points the people and the technology of the system toward a common goal. Coherence is what turns an advanced capability into an integration a system does not merely tolerate. A coherent change is self-reinforcing, while a change that has been bolted onto the system will eventually be counteracted.

Incoherence is the result of the work of many, and that is part of what makes it persist. Each element of the system was optimized in isolation. Coherence is a result of design, and is not produced automatically by a structure. This is why the next meaningful advancement should not be seen as developing yet another tool. Rather, it is a shift in the mindset of leaders in regard to what they are constructing. The focus becomes integrating technology in the service of the purpose of the organization, rather than merely implementing the technology.

From Efficiency to Wholeness

The framing of conversations around artificial intelligence (AI) in health care is shifting, albeit slowly. The framing started around efficiency, saving time, and automation. It was the most reasonable framing and the starting point, but was never the complete picture. Perhaps the most interesting potential is that AI can add back to the meaning of the work by removing the cumbersome tasks that crowd the human element of care, restoring the unencumbered element of care to the clinician.

That is not an overly emotional statement. It is the reality of the situation. When workflows are constructed around the needs of the patient, and the technology and data are used to assist the clinician, rather than dictating and restricting, positive change occurs. The clinician finds relief. The patient receives care from a clinician who is present and engaged, rather than from a clinician who is distracted and buried in their technology. AI is not the source of the compassion in the example. AI is a tool that can be used to restore the compassion and the human element to care that is meaningful.

Designing for It on Purpose

None of these positive changes will happen by accident. This will require purposeful design. This places an emphasis and a specific responsibility on leadership. AI cannot be a project that is distant from the core mission of the organization and technology. AI needs to be a tool that is integrated in the processes of decision making.

Practically speaking, this entails several steps. First, we must implement ethical oversight as part of the everyday operational decision-making process. This oversight will not be seen as a review we pass at the end of the process. Second, we must recognize, measure, and act upon the indicators of health and wellbeing within the workforce as the earliest warning signals. These include the time returned to people and the trust between team members and the tools they utilize. Third, we must assume that ethical explainability is built within the tools we create, rather than treating ethical explainability as a feature of a product we will add if time allows. These are not boxes to check. They are the difference between an organization that uses AI and one whose technology and work have actually come together.

The Human Core

At the center of all of this is the belief that the primary measure of the success of AI in the healthcare industry is not in the models, nor is it in the accuracy scores. It is people. It is people being able to feel that their work is now safer. It is people being able to feel that finally the technology is supporting the mission.

Once those conditions have become actualized, the system truly starts the healing process, and not due to some sort of automation. It begins healing because it has achieved a level of coherence, and the previously discordant components are now working collaboratively. Making healthcare faster will not bring coherence. Helping it hold together can, and that is the aspect of this effort that is worthy of being pursued, where the data, models and workflows are pointed at what should always be the ultimate goal: to provide better care.

Christopher Hutchins Founder and CEO, Hutchins Data Strategy Consultants