Who Owns Artificial Intelligence?
Ask four executives in the same organization and you will get four answers. Not because they disagree, but because the agenda is attached to people rather than to structures, and the people are moving.
By RaeAnn, Founder and Chief Executive Officer, HLTHWORKS
An ordinary career move, and what it reveals
In October 2025, Longitude Health named Vishal Agrawal, M.D., as Chief Executive Officer. Longitude Health is a collaborative founded in 2024 by leading not for profit health systems to build shared solutions for healthcare administration. Its board is chaired by the chief executive of one of the largest systems in the country. Among the stated priorities under the new leadership is the application of agentic and predictive artificial intelligence to support clinicians and streamline operations.
He arrived from Humana, where as Chief Strategy and Corporate Development Officer he had carried responsibility for corporate strategy, mergers and acquisitions, and the design of a new operating model and member experience.
This is an ordinary and entirely legitimate career move. It is worth studying precisely because it is ordinary. An executive who held the operating model agenda at a national payer now leads an artificial intelligence agenda on behalf of a group of health systems. The mandate did not simply move between organizations. It crossed the payer and provider boundary, which is the one place in this sector where nobody owns the ground.
Multiply that by the current rate of senior movement in healthcare. Segment presidents with decades of tenure are retiring. Transformation mandates are graduating into operating roles or concluding. Chief digital, chief information, chief strategy, and chief analytics roles rotate on their own cycle, and increasingly the replacements arrive from outside the sector entirely, from technology companies and from retail.
That churn is not the problem. Organizations survive executive transitions routinely, and always have. The problem is what happens to an artificial intelligence agenda that was attached to a person rather than a function.
Artificial intelligence in healthcare is owned by individuals rather than by structures. And the individuals are moving.
What leaves when the executive leaves
Consider what actually departs with a senior leader who has been the center of gravity for artificial intelligence.
- The inventory, if it existed at all, which typically existed in that leader’s head or in their team’s working files rather than as an institutional record.
- The reasoning behind which tools were approved and why, including the conditions attached to each approval.
- The informal understandings with vendors about what would and would not be done with the organization’s data.
- The knowledge of which deployments were pilots and which quietly became production without anyone re-approving them.
- The judgment about which tools were assistive and which had drifted into acting on their own.
A successor inherits the deployments and none of the reasoning. Within a quarter they are operating an artificial intelligence estate they did not build, cannot fully enumerate, and did not approve, and the organization holds no record that would allow them to reconstruct it. They then face a choice between ratifying everything they found and stopping things they do not understand. Both are expensive and neither is governance.
WHY THIS IS WORSE IN HEALTHCARE THAN ELSEWHERE
Three conditions compound the problem in this sector specifically.
- The obligations outlast the executives. A regulator examining a decision made in 2025 does not adjust its questions because the person who made it has moved on. The organization answers, not the individual.
- The tools touch care and payment. An undocumented deployment in most industries is an operational inconvenience. In healthcare it can sit in the path of a clinical determination, a bill, an authorization, or a denial.
- Adoption outran documentation everywhere at once. Nearly every organization deployed faster than it governed, which means the reasoning gap is not an isolated failure at one company. It is the sector’s default condition.
The three answers organizations give
Ask who owns artificial intelligence and you will hear one of three answers. Each is defensible. Each fails in a different way, and none of them is ownership, because ownership means one person can be held to a result.
|
THE ANSWER |
WHY IT IS GIVEN |
HOW IT FAILS |
|
Everyone owns it |
A committee with broad membership and a regular meeting. It looks inclusive and it distributes political risk. |
A committee cannot be held to a number and cannot be replaced when it underperforms. Broad membership without authority to stop a deployment produces consultation rather than governance. |
|
Technology owns it |
The tools are technical, the platform is technical, and the people who understand them sit there. |
It holds while the question is infrastructure. It breaks the moment the question is whether a tool should be used at all, which is a business, clinical, and regulatory judgment rather than a technical one. |
|
The most enthusiastic executive owns it |
Usually whoever is moving fastest and is measured on delivery. Frequently the transformation or innovation leader. |
Fastest adoption, weakest governance, and total dependence on one person. This is the combination least likely to survive a transition, and the one most common today. |
THE STRUCTURAL CONFLICT NOBODY HAS RESOLVED
In most organizations the executive sponsoring the largest volume of artificial intelligence deployment has no accountability for governing it, and is measured on speed of delivery. Governance and speed pull against each other, and when they sit with different people, the person measured on speed wins every quarter. This is the mechanism that produces an inventory nobody can produce on request, tools that arrived without approval, and a program with no record of ever having stopped anything.
An office measured on delivery, sponsoring the fastest moving technology in the sector, with no accountability for its governance and no authority to stop anything, is not a governance function. It is an acceleration function.
The agenda is leaving the organization entirely
There is a second movement underway, and it has received almost no governance attention.
Consortia, shared services entities, and jointly owned platforms are increasingly the vehicle through which health systems pursue administrative and artificial intelligence transformation. The logic is sound. No single organization can fund the work alone, the problems are common across all of them, and pooled scale produces solutions none of them would build individually.
But it raises a question most participating boards have not asked. When an organization joins a collaborative that will develop and deploy artificial intelligence on its behalf, whose governance applies?
The consortium sets the direction. The member organization carries the regulatory obligation, the patient relationship, and the liability. A board that approved membership without asking how oversight travels between the two has approved a capability and inherited an exposure.
FOUR QUESTIONS BEFORE JOINING, OR AT THE NEXT RENEWAL
- Who validates what the collaborative builds, and against whose population and data?
- When the collaborative deploys into our environment, does the tool enter our inventory, our risk tiering, and our monitoring, or does it sit outside all three?
- If our governance standard and the collaborative’s standard differ, which one governs, and who decides?
- If we exit the collaborative, what happens to our data, to any model built or improved with it, and to the tools already running in our environment?
The same four questions apply to a delegated vendor, a management services organization, or a private equity sponsor deploying a shared platform across a portfolio. Anywhere the capability is built outside the organization that carries the obligation, the governance has to be written down or it does not exist.
The structural answer
The organizations that have resolved this did not resolve it by choosing a better executive or by winning the ownership argument. They resolved it by attaching the artificial intelligence agenda to artifacts that survive a departure.
- An inventory that exists as a record rather than as knowledge. If the person who knows what is running is the only place that information lives, the organization does not have an inventory. It has an employee. The record must carry every tool: purchased, built internally, and introduced through a partner or collaborative.
- Written classification and approval reasoning. Not only what was approved, but why, by whom, and under what conditions. A successor who can read the reasoning can defend the decision or revisit it deliberately. One who cannot will either ratify everything or stop everything.
- A single named accountable executive, with the role defined independently of the person in it. The definition should survive three occupants. If the mandate must be renegotiated with each appointment, it was never a mandate. Name the person, write the accountability down, and attach budget to it.
- Independent review on a schedule. The only mechanism that produces an accurate account of the estate on a date certain, regardless of who occupies the chair that quarter, and the only one that reaches tools arriving through a consortium or shared services partner rather than through your own procurement.
None of those four requires resolving the ownership debate philosophically. They make the debate less consequential, which is both more achievable and more durable. An organization with those four artifacts can lose any single executive and still answer a regulator, a board, or a plaintiff attorney within days.
TWO BOUNDARIES THAT HOLD REGARDLESS OF WHO OWNS THE AGENDA
The party that deploys cannot be the party that reviews. Compliance and internal audit attached to the function delivering artificial intelligence produces clean reports and no assurance.
Nobody gets to bypass procurement for speed. Nearly every tool that entered an organization unrecorded entered because someone senior judged the process slower than the opportunity. That judgment is almost always made by whoever is measured on delivery, and it is the single largest source of the tools nobody can account for.
Five questions for the next C-Suite or Board meeting
Written to be asked without technical preparation, and to be answerable in a sentence by the executive who owns each one.
- How many artificial intelligence tools are in production here, and how confident are we in that number? A stated confidence below one hundred percent is the credible answer. One hundred percent means nobody has looked hard.
- Who is the single named executive accountable, and is it in writing with budget attached? Three executives each holding part of it means none of them holds it.
- If that executive left next month, what would leave with them, and how long to reconstruct it? This is the question this article exists to ask. Nobody enjoys answering it and everybody learns something.
- Where we have joined a collaborative or shared services entity, whose governance applies to what it builds for us? The consortium sets direction. We carry the obligation.
- Who reviews this independently of the people who built and deployed it, and when did this board last see that review? If the gap between the last review and the last board sighting is long, the gap is the finding.
A closing observation
Healthcare organizations have spent three years asking who should own artificial intelligence and treating it as a question about org design. It is not. It is a question about continuity.
The executives currently holding these agendas are capable, and many of them are doing the work well. They will also move, retire, be recruited, or be reorganized, because that is what senior executives do and always have. An agenda that cannot survive that is not an agenda. It is a dependency.
The organizations that will be defensible in three years are not the ones that answered the ownership question most elegantly. They are the ones that wrote enough down that the answer stopped mattering so much.
Sources
- Consortium leadership. Longitude Health appointment announcement, October 2025; Becker’s Hospital Review; Modern Healthcare; Newsweek coverage of the same appointment.
- Prior role and scope. Humana published leadership biography and executive appointment announcements.
- Executive movement in the sector. Humana insurance leadership transition announcement, December 2025; Centene Corporation leadership announcements, January and December 2022; Healthcare Dive and Becker’s Payer Issues coverage of both.
- Oversight and independence principles. McKinsey and Company, The role of the chief transformation officer; The role of the transformation office.
Organizational details reflect published information available at the time of writing and change frequently. Verify current titles and reporting relationships before relying on any specific reference. This article is provided for general informational purposes. It is not legal advice and does not describe the duties of any director or officer under the law of any jurisdiction.
CONTINUE READING
Article 9: The Chief Transformation Officer. A companion article in The Standard defining the role across health systems, medical groups, health plans, and administrators, including what the office owes each part of the enterprise and what it must never absorb.
Under Resources on the hlthworks.com The Preliminary AI Maturity Assessment. Thirty five items, seventy points, a self scored position in about twenty minutes.
Under Resources on the hlthworks.com Twelve Questions Every Healthcare Board Should Ask About AI. A governance brief written to be handed to an audit committee without a covering explanation. All available at hlthworks.com.
HLTHWORKS builds artificial intelligence governance, scalability, sustainability, oversight, audit readiness, and a mature enterprise solution for medical groups, health systems, health plans, payers, and the vendors, plus investors serving them.
Related Posts
