Healthcare AI Survey: Deployment Is Outpacing Governance - What 133 Healthcare Companies Reported

Across health plans, health systems, physician groups, and the vendors serving them, deployment is running at roughly double the rate of strategy. Nearly a third do not measure return. Four in ten report that no quality measure has improved yet. None of that is a reason to slow down. It is a reason to build the foundation the deployment is standing on.

By RaeAnn, Founder and Chief Executive Officer, HLTHWORKS | Published from RISE West, San Diego

The findings

We surveyed the healthcare market ahead of RISE West and received 133 responses. Every respondent answered every closed question. Health plans made up roughly forty-two percent. The balance were health systems, physician practices, accountable care organizations, vendors, and consultants, which matters when reading the numbers and is addressed below.

43.6%

MOVEMENT

21.1%

GOVERNANCE

30.8%

DO NOT MEASURE RETURN

39.1%

NO MEASURE IMPROVED

The first two numbers are the article. Forty-three point six percent report movement, meaning they are scaling artificial intelligence across functions or operating enterprise-wide. Twenty-one point one percent report governance, meaning a defined artificial intelligence strategy tied to business performance. Movement is running at roughly twice the rate of governance, and sixty-three of 133 report no formal strategy at all.

That is not a gap between leaders and laggards. It is a gap inside the same companies. The organizations scaling fastest are frequently the ones with the least defined strategy underneath the scaling, which is a different and more uncomfortable finding than a market divided into fast movers and slow ones.

Five things came through clearly.

Movement is outpacing governance, at roughly two to one. This is the finding everything else in the survey sits on top of.
Scale determines maturity almost perfectly. Confidence is not optimism in this dataset. Among healthcare companies deployed enterprise-wide, forty-six percent report very high confidence that AI will improve quality ratings. Among those still piloting, thirty-five percent report low or no confidence. Confidence is earned through execution.
Health plans are more skeptical than the vendors and consultants who share the sample. That gap is consistent across questions and is the single most useful reason to read the health plan subset separately.
Measurement discipline arrives after deployment rather than before it. Thirty-one percent do not formally measure return. Fifty-six percent say it is too early to evaluate whether the investment met expectations.
The market is spending on quality ratings while naming risk adjustment as the larger collaboration opportunity. Investment and stated opportunity are pointed at different places.

A note on the sample, because it changes how the numbers should be read. Vendors, consultants, and California respondents are each overrepresented relative to the underlying healthcare market. The sample is not nationally representative, and seventeen of the 133 records carry a response quality flag and are retained in the totals. The health plan subset of fifty-six is the more conservative read.

Confidence in this dataset is earned through execution, not assumed at the outset.

What it means

The instinct on seeing these numbers is to argue that the market should slow down until governance catches up. That is the wrong conclusion and nobody is going to follow it anyway.

The right conclusion is narrower. Healthcare companies are deploying artificial intelligence on top of a foundation that was not built for it, and the four numbers on the previous page are all symptoms of the same underlying condition. You cannot measure return on a system you have not inventoried. You cannot attribute a quality improvement when the data moved three times between the model and the measure. You cannot govern what six different business units bought separately.

Which reframes the problem. This is not an artificial intelligence maturity gap. It is a data, security, and accountability maturity gap that artificial intelligence has made visible and expensive.

Four implications follow, and each one has a number attached to it.

MEASUREMENT REQUIRES LINEAGE, NOT DASHBOARDS.

Thirty-one percent do not measure return and thirty-nine percent report no measure improved. In most healthcare companies that is not indifference. It is that the data moved through so many hands between source and outcome that attribution is genuinely impossible. Return becomes measurable when lineage exists, and not before.

GOVERNANCE FOLLOWS THE INVENTORY.

Sixty-three healthcare companies report no formal strategy. A strategy written before anyone has counted the systems in production is a document rather than a control. The sequence is inventory, then classification, then strategy, and it does not work in the other order.

THE SILO IS THE COST DRIVER.

When each function procures its own tool, the company pays for the same capability several times, cannot compare performance across them, and multiplies its vendor exposure. The enterprise-wide companies in this sample are not more confident because they are braver. They are more confident because they can see across the whole thing.

SECURITY AND ACCESS ARE THE LIMITING FACTOR, NOT MODELS.

Every capability described above depends on data that can move safely between functions. Where it cannot, the company defaults to point solutions, which is how it arrived here. Access architecture determines what is possible before any model is selected.

What we are working on

These findings describe the work HLTHWORKS does, so it is worth being direct about it rather than leaving it implied.

We are not selling a model. We build the foundation underneath one, in five places.

Inventory and classification. A complete accounting of every artificial intelligence and automated decision system in production, including capability embedded inside vendor products that was never procured as artificial intelligence. Most inventories we encounter are between fifty and seventy percent complete, and the missing portion is almost always vendor-embedded.
Data structure and lineage. Reducing the number of times data shifts hands between source and outcome, and documenting the shifts that remain. This is what makes return measurable and what makes an attribution claim survive examination.
Security and access architecture. The permissions, controls, and data movement rules that let information cross functions safely. Without this, silos are not a cultural problem. They are a security necessity, and no amount of governance language dissolves them.
Independent audit and assurance. An outside read of the position, with evidence tested rather than described, delivered as a board-ready finding. Self-assessment is a starting point, not evidence, and independence is the first thing an external reviewer examines.
Enterprise scaling across silos. Designing capability once and deploying it across functions, rather than procuring it separately in each. This is the difference between the enterprise-wide companies in this sample and everyone else.

None of that is exciting, and all of it is what separates the forty-six percent who are confident from the thirty-five percent who are not.

The read

A market that deploys faster than it governs is not behaving irrationally. It is responding to real pressure with the tools available, and the pressure is not going to ease.

But the healthcare companies in this sample that report the highest confidence are not the ones that moved fastest. They are the ones that can see across their own enterprise, measure what they deployed, and answer for it. That capability was built deliberately, before it was needed, by people who were not rewarded for building it at the time.

The gap between forty-four percent deploying and twenty-one percent with a strategy will close. The only question is whether it closes because governance caught up, or because something went wrong loudly enough to force it.

About the survey

The RISE AI Industry Benchmarking Survey was fielded ahead of RISE West 2026 and captured 133 responses across twenty closed questions and one open-ended question. All respondents answered every closed question. Seven questions permitted multiple selections, so those percentages are expressed as a share of all 133 respondents and sum to more than one hundred percent.

Respondents represented eighteen company types across thirty-five states, spanning health plans, health systems, physician practices, accountable care organizations, vendors, and consultants. Health plans, comprising Medicare, Medicaid, commercial, and multi-line, totaled fifty-six respondents or 42.1 percent. Twenty percent of respondents sat at the C-suite, executive vice president, or senior vice president level. Risk adjustment and quality improvement together accounted for roughly one quarter of respondents by job function.

Several questions in the instrument used Medicare Advantage and Star Ratings framing, reflecting the conference at which the survey was fielded. Respondents spanned multiple lines of business, so those results are reported here in the broader terms respondents actually operate in. The sample is not nationally representative. Vendors, consultants, and California-based respondents are each overrepresented relative to the underlying healthcare market. Seventeen of the 133 records carry a response quality flag and are retained in the totals reported here. Conclusions drawn from the full sample should be read alongside the health plan subset.

The complete findings and analysis, including cross-tabulated trends and open-ended response themes, is available at hlthworks.com.

Also in The Standard

Article 14. Nothing to Inspect. Why compliance cannot see what it is now accountable for.

Article 15. AI in Health Plan Member Experience Is a Risky Venture. The who, what, where, and why.

Resources. The AI Regulatory Update. The Preliminary AI Maturity Assessment. Twelve Questions Every Healthcare Board Should Ask About AI.

HLTHWORKS transforms Medicare Advantage, Commercial, and Medicaid health plans, driving efficiency in the business of health, impact in the value and quality of care delivery, and simplicity in the patient journey.