To say that today’s medical practice faces challenges from several different fronts — care quality, patient experience and satisfaction, staffing, burnout, access, and financial sustainability — is an understatement. On top of that, technological change can be hard to keep up with, let alone maintain any consistency while meeting patient needs. Enter artificial intelligence (AI): Will this reduce or introduce complexity to care delivery? This article suggests an approach that leads to stability and growth for medical practice leaders: addressing the disruption and innovation through a systematic approach to daily and long-term issues to build a model for success.

AI can be found — to a limited degree — in practice management (PM) and EHR systems, as well as internet-based tools such as chatbots, scheduling software, and financial analysis. Who hasn’t tried ChatGPT, Gemini, Copilot or another tool to research answers? Which vendor hasn’t integrated some form of AI into its platform? Today’s PMS/EHRs are built for compliance and billing, and not for the full potential of generative AI. Thankfully, there are agents such as Abridge (ambient listening) and Inbox Health (patient billing) that integrate with existing systems. The future will bring further integration and fuller use of AI’s potential.

A word of caution: is the agent’s approach to the solution consistent with your processes? Does it meet privacy, functional and ethical standards aligned with your internal policies and federal/state regulations? Is your practice ready to use AI well? As Yuval Harri, author of Nexus, has suggested, AI can seem like “alien intelligence” —not artificial so much as the product of machine learning (ML) — which is a good reminder to set expectations.

Let’s think about your practice. How efficient is the flow of patients on a regular basis? How effective is the data from the first appointment call through billing, and how does patient compliance factor in? Will all this lead to a care plan that yields a positive patient experience and healthy outcomes? What can you do now? Where are the barriers or gaps you could fix even without AI? How will you integrate AI into those processes when you’re ready?

Start with the fact that anything new from your technology stack must fully integrate with what you do now. Do you have process maps for your key processes? Have you reviewed them to find barriers or gaps? What is your current process to manage change or disruption? What are the current skills of the staff, and what is their level of knowledge on the process — why, how, and when they do their “routine” work? What can be done to re-skill or upskill them to be effective and efficient in the new workflow?

To integrate changes from your vendors or your own AI-generated ideas, there will be disruption to the current flow. In Beyond Disruption, W. Chan Khan and Renée Mauborgne describe two paths. First, disruptive creation, which can lead to job loss — a common fear as AI becomes more prevalent. Second, nondisruptive creation, in which disruption occurs without job loss. Leadership must determine which path you’re on. Can you achieve improvement with existing staff, using change management techniques and reskilling/upskilling to succeed? Don’t design everything in the C-suite — use the “brains” of the staff to diagnose what will be disrupted. As you analyze a process, ask what brings value to the customer and what is non-value activity that should be eliminated.

Keeping with the theme of disruption, notice that Kim and Mauborgne pair it with creation. It’s not enough to change; you must also innovate. Innovation requires a proactive mindset open to a new way of doing things and tolerant of small failures. It happens when you challenge existing processes and the assumptions behind them to discover a new way to perform the task. The goal is to create a new and better way to bring value to the customer — and “customer” may be a colleague or provider, which ultimately improves value to the patient.

A recent exchange with your vendor convinces you their approach to AI is great and should be implemented immediately, starting with patient scheduling. You green-light it at the front desk. Training occurs, the team is ready, and you go live. Suddenly, schedules glitch for every provider, and the MAs don’t understand what is happening. Talk about disruption!

What happened? You treated a silo, not a system. A change in one aspect of the patient process affects many others. Systems thinking recognizes the interconnectedness and relationships across all activities. It’s time to pause and acknowledge the structural issue; otherwise, a fix in one area can create greater disruption elsewhere. A great idea gone wrong.

To address this, look at the end-to-end process that delivers a great patient experience. It’s not built by silos; it’s an integrated set of activities. A simple process map begins with the pre-visit activities (e.g., scheduling, eligibility verification), continues through the visit (check-in, triage, clinician interaction, post actions, and checkout), and finishes with post-visit work (care plan compliance monitoring and the revenue cycle). Better yet, think in three stages — pre-visit, visit and post-visit. See Figure 1 for a straightforward process map of each stage. Each stage stands alone, yet each impacts the others.

Figure 1 combines three process maps into one picture. These are simplified maps show the basic tasks related to pre-visit, visit, and post-visit stages of the overall patient experience, arranged on a single slide. You don’t need high tech to build this. You need staff input and sticky notes on a whiteboard or brown paper. Take a picture when complete to memorialize the effort, then transfer to a digital format if desired.

Figure 2 offers another tool that explains interconnectedness: the casual loop diagram. It highlights the importance of accurate demographics, insurance and relevant medical information in the pre-visit stage — and how that accuracy can positively (or negatively) influence the visit and post-visit stages in a single patient’s experience.

Within the process map, drill down with employees and ask simple questions. Why are you doing that step? The answer should be “to gather necessary information” or “to complete a required step in delivering care.” Then ask what are you doing, and how do you do it — demonstrate the clicks and handoffs. All too often the answer is, “That’s how Peggy wanted it done.” Problem: Peggy retired two years ago. Once you establish the current state, ask whether it could be done differently. This is where AI options come into play. Does an agent from your PMS/EHR apply? What benefits would it bring? Can you innovate further by integrating AI and eliminating non-value steps? Also ask: who uses the output and what happens if it’s wrong? These small questions, answered by people who do the work, surface specifics that leaders can’t see from a distance.

You may identify a task that truly should be disruptive — and will improve performance — but you implement it without recognizing its impact elsewhere. That’s a myopic view. Systems thinking forces the broader view.

There are some basic systems-thinking steps to develop a more comprehensive understanding of disruption that has (or may) occur. In today’s busy practice, pressured to “make an improvement,” we too often minimize the most important step: clearly define the problem. Take time to identify and set boundaries for the target.

Next, map the actions/steps in the entire process with the people who do the work. You may not need a polished picture to start, but tracing the steps needed to reach the desired outcome will reveal weak areas — especially where a fix would have the biggest impact. Reviewing performance over time may reveal patterns a single snapshot missed. Fix the biggest issue first, then revisit the process later to fix the next biggest issue — the essence of continuous process improvement (CPI).

At this point, begin to develop the solution. Involve the team impacted by the issue. When implementing a transition to a new way of working, having those who will live with the solution help design it makes adoption faster and more effective. It is one thing to fix the immediate issue and smooth the disruption; it is another to broaden the thinking to eliminate and innovate. Try a new way of doing things. If the issue can be fixed by an agent within the PMS/EHR — or by something you have discovered through AI — all the better.

Design and implementation are not the final step in systems thinking. The busy manager might assume all is well once a solution is in place. Instead, build monitoring and follow-up into the project. Expect to make adjustments as the improvement evolves, and be prepared to fix more than one thing.

Taking a systems perspective in implementing AI agents is essential. Think of the Titanic and its run-in with an iceberg — a major disruption to a smooth sail. An iceberg reveals only its tip; roughly 80% of its mass is unseen. Focus on the whole, not just what is obvious.

From another perspective, an iceberg is amazing to see up close. Its beauty above the waterline is impressive; looking below reveals even more. An iceberg is not only a disruption; it also offers a way to think about possibility. As we reflect on AI, many future applications are not yet visible. Preparing now — through a systems perspective — will lead to a successful practice that delivers expected care and a positive patient experience, with a satisfaction for all as you meet and exceed outcomes.