Hospice AI Governance Is Becoming a Leadership Responsibility

Artificial intelligence is moving rapidly into hospice operations.

Providers are using or evaluating AI to reduce administrative work, strengthen care coordination, improve access to information, support after-hours triage, analyze organizational data and review interactions with patients and families.

The potential is significant. But so are the governance challenges.

Two recent Hospice News reports highlight a tension every hospice leader should understand: while organizations are expanding their formal use of AI, employees may already be using consumer AI tools outside approved systems and policies.

This means AI is no longer only a technology project. It is an organizational governance responsibility.

The CIO’s role is changing

Hospice technology executives have traditionally focused on infrastructure, cybersecurity, EHR reliability, integrations and technical support. AI expands that role considerably.

Hospice News recently described how VITAS Healthcare and Empath Health are using AI across several operational areas.

VITAS reported using AI to accelerate software development, improve its internally developed EHR, support administrative and IT requests, and monitor call-center interactions for quality assurance and employee coaching.

Empath Health described using AI to reduce administrative burden, improve patient access, strengthen after-hours triage, support care coordination and allow employees to ask natural-language questions of organizational data.

These examples demonstrate that AI is moving beyond isolated experimentation. It is beginning to influence how hospice organizations communicate, make decisions and manage daily operations.

As a result, hospice technology leaders must increasingly function as business strategists and governance leaders—not only as technical experts.

They need to help the organization determine:

• Which business problems are appropriate for AI
• What information an AI system may access
• Where protected health information is processed and retained
• Whether customer data is used to train an external model
• Which outputs require clinical, compliance or management review
• How accuracy and operational value will be measured
• How errors, disagreements and overrides will be documented

These questions cannot be delegated entirely to an outside technology vendor. The hospice remains responsible for its clinical records, compliance processes and use of patient information.

The hidden challenge of shadow AI

Formal AI projects represent only part of the picture.

In a separate Hospice News report published September 23, legal professionals warned about “shadow AI”—employees independently using consumer AI applications without the organization’s knowledge or approval.

Examples can include using a public AI tool to draft a visit note, summarize patient information, improve a physician narrative or prepare communication for a family.

Employees may not have harmful intentions. In many cases, they are simply trying to reduce administrative burden and complete their work more efficiently.

However, this activity can create significant questions:

• Was protected health information entered into an unauthorized system?
• Does the application retain or reuse the information?
• Can the organization determine which content was written by a clinician and which was generated by AI?
• Was the output reviewed for accuracy?
• Does the final record reflect the clinician’s independent judgment?
• Can the organization reproduce the process during an audit or investigation?

Physician narratives and eligibility documentation deserve particular attention because they can influence hospice eligibility and are frequent areas of regulatory scrutiny.

An AI-generated statement that appears reasonable may still omit important context, introduce unsupported language or make the documentation look more uniform than the underlying clinical circumstances justify.

Prohibiting AI is not enough

A blanket prohibition may appear to be the safest response, but it does not address the reason employees turn to these tools.

If staff members are overloaded, struggling to locate information or repeatedly completing manual documentation tasks, the operational need remains even after AI is prohibited.

An effective governance strategy should therefore combine reasonable restrictions with approved alternatives.

Hospice organizations should consider five foundational steps.

1. Discover current AI use

Ask employees where and why they are already using AI. The objective should initially be discovery rather than punishment.

Leadership needs an accurate picture of the workflows driving unsanctioned use.

2. Establish an acceptable-use policy

The policy should explain which tools are approved, what information may be entered, which activities are prohibited and when human review is mandatory.

It should address consumer AI tools explicitly rather than relying on a general technology policy.

3. Provide controlled tools

Employees are less likely to use unauthorized applications when the organization provides a secure alternative that fits their workflow.

Approved systems should operate within clearly defined data, privacy and retention boundaries.

4. Preserve human accountability

AI can identify an exception, retrieve supporting information or prepare a draft. It should not silently make high-risk eligibility, billing or compliance decisions.

The responsible professional should review the evidence, accept or reject the recommendation and document the final decision.

5. Maintain provenance and an audit trail

The organization should be able to identify:

• What source information was reviewed
• What the AI generated or recommended
• Which rule or policy applied
• Who reviewed the output
• What decision the reviewer made
• Whether corrective action was completed

This converts AI from an opaque shortcut into a governed decision-support process.

A practical opportunity for private AI

For hospice providers, private AI offers an alternative between unrestricted consumer tools and complete avoidance of AI.

A customer-controlled private-AI environment can use the hospice’s approved policies, procedures and regulatory sources while limiting where sensitive information travels. It can help review documentation, identify missing evidence and organize exceptions without making the final clinical or compliance decision.

Privacy alone, however, is not sufficient.

A well-designed private-AI system must also provide source-linked findings, defined workflow boundaries, role-based access, human approval, audit logging and an effective way to challenge or override an AI recommendation.

The goal should not be autonomous hospice compliance.

The goal should be to help qualified people review more information consistently, locate supporting evidence faster and make better-documented decisions.

The leadership question

The most important AI question for a hospice is no longer, “Should we use AI?”

AI may already be present through formally approved platforms, embedded vendor features or employees’ independent use of consumer tools.

The better questions are:

Where is AI being used? What information can it access? Who is responsible for reviewing its work? And can the organization explain the final decision?

Hospice organizations that answer those questions now will be better prepared to benefit from AI without sacrificing privacy, compliance or human judgment.