AI + Healthcare, Part 3: AI’s Next Phase Is About Control, Not Speed

The first two pieces in this series examined two problems: AI may develop treatments faster than insurers can pay for them, and approved products may keep changing after regulators clear them.

The next issue is control.

AI companies have spent years competing on speed and capability. Now investors, enterprise customers and health systems are asking a more practical question:

Can this technology be governed, protected and turned into measurable value?

Recent developments show the technology industry beginning to wrestle with the limits of acceleration. Public debate around frontier-model development is becoming more pointed, while Microsoft has proposed limits on how its AI models may be used even as it continues to invest heavily in the technology.

CNBC on Anthropic CEO Dario Amodei
CNBC on Microsoft’s AI model limits

This is not necessarily a retreat from AI. It is a move from “How fast can it improve?” to “Can organizations use it safely and reliably?”

Control is becoming part of the product

Enterprise buyers want straightforward answers. What happens to their data? Can the system’s output be audited? Who is responsible when the model changes? Can its use be limited? Will the vendor provide notice before a major update? Can the company show a return on the investment?

Those questions are becoming part of the product itself.

A less powerful system that can be monitored and integrated into a company’s workflow may be more valuable than a more advanced system that creates uncertainty around data, liability or reliability. Technical capability still matters, but capability without control can become a business problem.

For healthcare organizations, that distinction is even more important.

Healthcare cannot treat every update as routine

AI-enabled products may change through new data, software releases or revised algorithms. Some updates fix technical issues. Others may alter recommendations, intended use, patient population or risk profile.

Leadership needs a process for evaluating and documenting those changes before release.

FDA authorization does not resolve every post-approval question. A hospital may ask whether an update changes its workflow or liability exposure. A payer may want new evidence. Providers may need additional training. Contracts may need to address version control, cybersecurity and responsibility for errors.

The faster a product changes, the more important its governance becomes.

Investors will look beyond the model

Technical performance will remain important, but it will not be enough. Investors will also want to know whether a company can protect customer data, maintain reliable performance, demonstrate outcomes, manage regulatory obligations and control deployment.

They will also ask whether usage can become recurring revenue.

That makes governance part of the company’s infrastructure and valuation story. A company that can explain how its system is monitored, updated and protected may be easier to finance, partner with and scale.

The winners may not be the companies that move fastest in every direction. They may be the companies that understand where speed creates value and where discipline protects it.

Five questions executives should ask

Before deploying an AI system, leadership should be able to explain what decision the system is helping someone make. That sounds simple, but it forces the organization to define the system’s purpose instead of treating AI as a general solution.

Leadership should also understand what data the system uses and who controls that data. That includes questions about ownership, confidentiality, access, retention and whether information is used to train or improve another system.

The organization should know how updates will be tested and documented. A software change may be minor, or it may affect performance, workflow or patient risk. The company needs a way to tell the difference.

Someone must remain responsible for the outcome. AI may assist with a decision, but responsibility cannot be assigned to the software. Contracts, internal policies and operating procedures should reflect that reality.

Finally, leadership should define what evidence will show that the system creates value. That may involve better outcomes, lower administrative costs, faster development, improved patient access or a more reliable workflow. Without a clear measure, adoption can become expensive experimentation.

These questions apply to drug development, diagnostics, clinical software, health-system operations and administrative tools.

They also affect the supporting contracts. Data rights, intellectual property, confidentiality, audit access, cybersecurity, indemnification, performance standards and termination rights all become more important when a product can change after it is deployed.

The next AI advantage

AI should continue improving. The market should continue testing what it can do.

The companies best positioned to succeed will understand that control is not the enemy of speed. Control is what allows speed to become something customers, regulators, investors and patients can actually use.

Lanton Strategies International helps healthcare and life-sciences companies understand the policy, reimbursement, market-access and commercial questions that determine whether innovation can move from development into sustainable use.

Lanton, Lanton & Sosa Law PLLC advises healthcare and life-sciences companies, technology businesses and other regulated organizations on the legal issues that arise as products and services enter the market, including contracts, data rights, intellectual property, cybersecurity, regulatory compliance, commercial arrangements and liability.

Together, LSI and the law firm help clients evaluate both sides of the same question: how to turn innovation into a product or service that can be deployed, protected, reimbursed and built to last.

Earlier in the series:

Part 1: What Happens When AI Develops Drugs Faster Than Insurers Can Figure Out How to Pay for Them?

Part 2: What Happens When the Product Keeps Changing After Approval?

This article is for general educational and informational purposes only. It is not legal advice and does not create an attorney-client relationship.