Field Notes Newsletter #5 - September 2026
AI Can Find the Problem, But Can Your Workflow Act on It?
By Chris Talbot, Account Executive, Healthcare Division at SMG3
Healthcare technology is advancing at a remarkable pace. AI can surface risks earlier, predictive analytics can identify patterns hidden in mountains of clinical data, and virtual nursing, mobile tools, and clinical communication platforms are creating new ways to support care teams. Across health systems, leaders are investing in technologies designed to make care more connected, efficient, and responsive.
But introducing innovation and realizing its value are two very different things.
For all the attention given to what new technology can detect, predict, or recommend, a more fundamental question often gets less attention: What happens next?
An AI model might recognize that a patient is deteriorating. An EHR may show that a patient is ready for discharge. A workflow engine can identify and route the next task.
Documentation is not motion.
But identifying the next step doesn't ensure that it happens.
Someone still has to receive the information. They need to know the task is theirs. They need the right tools, access, and context to respond, and enough capacity to act. And increasingly, that means having a working, properly configured device or endpoint in hand when and where the work needs to happen.
That gap between knowing what needs to happen and making it happen may be one of the most important challenges facing healthcare technology today.
I think of that gap as the execution layer: the point where information, people, devices, and ownership have to come together in real time to turn intelligence into action.
Because intelligence isn't the same as execution.
AI Makes Workflow More Important, Not Less
There is an understandable tendency to think that increasingly intelligent technology will solve many of healthcare's workflow challenges. In some cases, it will. But AI can also expose weaknesses that were already there.
If the handoff between identifying a problem and acting on it is unreliable, AI may simply identify that problem faster and more frequently. Better detection doesn't necessarily lead to better outcomes when the execution layer underneath it remains fragmented.
Before adding more intelligence to a workflow, healthcare leaders should therefore be asking some decidedly practical questions:
Who owns the next action? How does the task reach that person? Can they distinguish it from the dozens of other signals competing for their attention? Do they have a functioning device or endpoint, with the right applications and configuration, to respond? Can the unit safely absorb the additional work? And can the next person in the workflow see that the task is complete?
These questions become even more important as AI evolves from recommending actions toward initiating or assigning them.
An algorithm may know that a task needs to be completed. It may not know that a unit is short-staffed, that a nurse is responding to a more urgent situation, or that circumstances changed five minutes ago. That is why AI should support frontline judgment, not automate overload.
The Pilot Isn't the Finish Line
This execution gap also helps explain why promising healthcare technology pilots don't always translate into successful enterprise deployments.
Pilots often operate under ideal conditions: engaged clinical champions, carefully configured technology, additional IT support, vendor representatives nearby, and heightened attention from leadership. The real test comes later.
Does the workflow still work at 3 a.m.? Does it work for the float nurse who wasn't part of the pilot? Does it work across different facilities, shifts, devices, and staffing conditions without someone standing nearby to troubleshoot?
A pilot can prove that an idea works. Standardization and operational discipline prove that it can survive.
That distinction matters as health systems race to deploy AI and other next-generation technologies. Scaling innovation isn't simply about deploying more technology. It requires building a dependable execution layer around the people expected to use it, including the devices, applications, access, and support that allow the workflow to function under normal operating conditions.
From Innovation to Execution
Perhaps the next phase of healthcare transformation should focus less on asking, What else can this technology do? and more on asking, What work does this technology actually remove?
Does it reduce friction for clinicians? Does it make coordination easier? Does it clarify ownership? Does it make the right action easier to take? Can the workflow expand without requiring a proportional increase in people to keep it running?
Those aren't implementation details. They are increasingly the difference between technology that demonstrates potential and technology that creates lasting value.
Healthcare doesn't need fewer ideas. It needs a more reliable way to turn those ideas into repeatable daily operations.
Because innovation doesn't create value when the algorithm identifies the problem. It creates value when the execution layer.
From the Field
Notable conversations you need to hear
EP53 - AHIMA's Jennifer Mueller: Trusted Health Information
Health Information professionals may not always be visible at the bedside, but their work touches nearly every patient encounter. In Episode 53 of Beyond the Blueprint, Jennifer Mueller, Senior Vice President of Health Information, Career Advancement, and Academic Affairs at AHIMA, joins Keith Washington and Kristin Baird to explore why trusted health information is becoming even more critical as healthcare enters an AI-enabled future. The conversation looks beyond the traditional view of Health Information as simply managing medical records to examine its expanding role in data quality, information governance, interoperability, privacy, and patient identity. Jennifer also offers an important reminder for organizations racing to adopt AI: start with the problem you are trying to solve, establish thoughtful governance, and make sure the information powering the technology can be trusted. Listen to the full episode to hear why the Health Information workforce, and the integrity of the data it stewards, will be essential to building healthcare’s next chapter.
Sound Bites
The moments that made us hit rewind
EP54 - Communication Breakdown: Who Owns the Next Step?
What happens after AI identifies a problem and someone actually has to act on it? In Episode 54 of Beyond the Blueprint, Gregg Malkary is joined by Chris Talbot of SMG3 Partners, Ron Remy of Mobile Heartbeat, and Angus Douglas of GE HealthCare for a candid discussion about the often-overlooked execution layer between insight and action. The conversation explores why successful pilots don’t always translate to enterprise-wide success, how critical messages become lost among hundreds of competing signals, and why clinician trust can quickly erode when alerts generate more noise than value. The panel also examines the less glamorous realities that can make or break even the most sophisticated technology: from dead or misconfigured devices to unclear workflow ownership and inadequate change management. Listen to the full episode for a practical look at what it really takes to move beyond the pilot and build technology-enabled workflows that clinicians and entire health systems can depend on.
Sound Bites
The moments that made us hit rewind
EP55 - AMDIS Roundtable - Before the Digital Front Door: How AI Is Reshaping the Patient Journey
The patient journey increasingly begins before someone ever reaches a health system’s digital front door. In Episode 55 of Beyond the Blueprint, Gregg Malkary sits down with Dr. Minal Shah, Dr. Eve Cunningham, and Dr. Deepti Pandita for an AMDIS Roundtable exploring how ChatGPT, search, wearables, home monitoring, and other AI-enabled tools are already influencing how patients interpret symptoms, decide where to seek care, and prepare for clinical encounters. The discussion examines both sides of this shift: patients may arrive better informed and ready for shared decision-making, but AI-generated guidance can also contribute to anxiety, unnecessary utilization, delayed care, and unreliable conclusions. The panel explores how health systems can respond by thoughtfully incorporating AI into patient navigation and pre-visit workflows while maintaining strong governance, human oversight, and clinical judgment. Listen to the full episode for a timely conversation about how healthcare leaders can help ensure AI becomes a trusted partner in the patient journey, not another source of complexity for patients and clinicians.
Footnotes
Further reading for curious minds
Every great conversation leaves you with something to explore. Footnotes is where we collect the books, articles, research, podcasts, and other resources mentioned or recommended by our guests, along with a few additional references that add context to the ideas discussed. Think of it as a curated reading list for anyone who wants to dig a little deeper. You're welcome!
Further reading:
Jennifer discusses AHIMA’s newly released 2026 Patient Naming Framework, which provides standards for consistently capturing patient names, including considerations around government-issued identification, newborn naming, multicultural naming conventions, and system configuration. She connects accurate patient identification directly to interoperability and AI-driven patient matching: poor or inconsistent source data can undermine even sophisticated matching algorithms.
In EP53, Jennifer emphasizes that organizations should establish governance and accountability before turning on AI capabilities and should begin with the problem they are trying to solve rather than buying technology and hoping it solves a problem.
3.) AHIMA: “15 Smart Questions to Ask Healthcare Artificial Intelligence Vendors” - The article provides practical questions health information leaders can ask vendors about governance, maintenance, data sharing, auditing, security, regulatory compliance, and long-term management.
4.) “How to Strengthen AI Governance at Your Organization” — Journal of AHIMA - Published August 19, 2026, this article argues that data quality becomes even more important as AI capabilities grow and recommends incorporating data-quality reviews throughout AI procurement, implementation, validation, and post-deployment monitoring.
5.) Anthropic: “Introducing the Model Context Protocol” - Ron Remy discusses Anthropic’s Model Context Protocol (MCP), introduced in November 2024, and its potential to make data from different systems more accessible to AI. Anthropic describes MCP as an open standard for connecting AI systems to the data sources and tools they need, which is highly relevant to the panel’s discussion of breaking down data silos and enabling AI to work across healthcare datasets.
6.) “Improving Clinical Alarm Management: Guidance and Strategies” — Biomedical Instrumentation & Technology - The article defines an alarm flood as 10 or more alarms occurring within 10 minutes, a volume considered beyond what a human operator can effectively manage.
7.) AMA — 2026 Physician Survey on Augmented Intelligence - AMA’s recurring physician AI sentiment surveys note that physician sentiment has become increasingly positive as clinicians gain experience with tools such as ambient documentation and AI-enabled clinical decision support. The latest survey found that 81% of physicians now use AI professionally, more than double the 38% reported in 2023. It also explores physicians' attitudes toward patients using AI, which directly connects to the roundtable's discussion of patients arriving at appointments with AI-generated health information.
Upcoming Events
September 27–29, 2026 / Baltimore, MD
October 12-14 / Las Vegas, NV
October 26-28 / Las Vegas, NV
October 29–31 Chicago, IL
November 29–December 3, Chicago, IL
Comments