AI Is Changing Healthcare. Are Your Preparedness Plans Keeping Up?
AI can help clinicians interpret information, monitor patients, and make decisions faster. But what happens when an AI-generated result is wrong, or the information behind it has been deliberately manipulated? And what happens when that result moves into the next system, the next department, or the next clinical decision? For healthcare leaders, these questions belong in the conversation about AI adoption from the beginning.
A 2025 narrative review in Risk Management and Healthcare Policy explores this tension. The authors describe AI’s potential across diagnostics, patient monitoring, chronic disease management, and remote care. They also examine risks involving inaccurate outputs, data leaks, algorithm manipulation, deepfake misuse, and errors spreading between interconnected systems. Read the study.
The more your organization relies on AI, the more important it becomes to understand how care will continue when that technology is unavailable or cannot be trusted.
AI Introduces a Different Kind of Disruption
Traditional downtime planning often starts with a recognizable problem: a system goes offline.
AI introduces another scenario. The system remains available. Information continues to appear. Staff can still complete their workflows. Yet an output may be inaccurate, misleading, or influenced by compromised data.
An inaccurate AI output does not automatically indicate a cyberattack. AI can produce errors without malicious interference. But teams still need a clear process for recognizing questionable information, escalating concerns, and deciding whether to continue using the tool while the cause is investigated.
That makes AI readiness a shared responsibility across clinical care, cybersecurity, IT, patient safety, and operations.
Connections Can Extend the Impact
The review highlights the risk of errors propagating when intelligent systems exchange information. For preparedness teams, that is a reason to look beyond individual applications and examine the workflows around them.
Consider a hypothetical AI-supported imaging workflow. Automated analysis helps prioritize studies for review. Staff identify questionable results, and the organization suspends the tool while investigating.
The immediate concern is the reliability of the analysis. The operational questions follow quickly:
Which studies need another review?
How will the team prioritize work without automated support?
Can existing staffing absorb the additional workload?
What delays should other departments anticipate?
Who will authorize the return to normal use?
Now consider where connected medical devices fit into that picture. Monitoring equipment, imaging systems, and other devices may supply information to AI applications or depend on connected software and services. Your plans should identify those relationships and establish what remains usable if a component is isolated or becomes suspect.
The response needs to account for the entire service patients depend on.
Move AI Risk Into Your Continuity Planning
Our recommendation is to start with one important clinical or operational workflow that uses AI. Bring together the people who use it, support it, oversee its risks, and manage the services that depend on it. Work through four priorities:
Understand the dependency. Identify what the AI tool does, what information it receives, where its outputs go, and which decisions it influences. Include vendor platforms and connected equipment in the discussion.
Define when to escalate. Give staff a clear way to report unexpected behavior or questionable results. Clarify who evaluates the concern and who has authority to pause use.
Build a workable alternative. Determine how the service would continue without AI support. Address staffing, equipment, documentation, capacity, and communication needs.
Plan the return to service. Establish who validates readiness, how affected work will be reviewed, and how teams will manage any backlog or reconciliation.
These steps turn a broad concern about AI into specific decisions your organization can prepare to make.
Practice Uncertainty Before an Incident Creates It
A useful exercise does not need to begin with a confirmed cyberattack.
Start with a suspected compromise involving an AI-enabled application or connected device. The scope is unclear. The vendor is investigating. Patient care must continue.
Ask participants to work through the first hour, the next shift, and recovery. Test whether teams can identify affected workflows, communicate restrictions, activate alternatives, and make coordinated decisions with incomplete information.
The review considers emerging technical safeguards, including blockchain, but it is a narrative synthesis rather than a test of a particular intervention. For organizations applying its lessons, a practical starting point is to examine their own dependencies and readiness gaps.
At Wakefield Brunswick, we help healthcare organizations do that through Business Impact Analysis, continuity and downtime planning, and realistic cybersecurity exercises. Our integrated approach connects clinical, operational, and technology teams around the decisions required to sustain care during disruption. Learn more about how we work
Is your organization ready for an AI disruption?
Ask yourself three questions:
Do your teams know how to recognize and escalate a questionable AI result?
Could critical services continue if an AI tool were paused today?
Does someone have clear authority to decide when it is safe to resume?
If any answer is uncertain, now is the time to test it.
Wakefield Brunswick builds realistic exercises around your AI-enabled workflows and connected devices. Your teams can practice making decisions with incomplete information before an incident forces them to.
