In 2012, MD Anderson Cancer Center began working with IBM on one of the most ambitious experiments in healthcare AI. The premise was compelling: combine the knowledge of a leading cancer center with Watson’s computing power and help physicians make better treatment decisions.

Five years and roughly $62 million later, MD Anderson allowed the contract to expire before Watson had been used to treat an actual patient. A university audit documented procurement problems, cost overruns and delays. The Journal of the National Cancer Institute also described the challenge of assimilating Watson into a hospital environment where important information lived in physician notes, medical shorthand and electronic records that the system could struggle to interpret.

I take a broader lesson from that history. Intelligence can be impressive in isolation and still create little operating value when it cannot understand the information around a process, participate in the work and hand the next action to the right system or person.

The same operating challenge has returned with agentic AI. Healthcare processes can cross clinical information, coverage rules, providers, payers, care teams and multiple operating systems. As CIO has recently noted in its coverage of enterprise architecture for agentic AI, traditional interfaces can move data between systems without supplying all the business context an agent may need to operate across them. In my experience, that is where the architecture problem begins.

The problem between the systems

Prior authorization is a useful example because it appears simple from the outside. A request is submitted and a decision comes back. Inside the enterprise, the process may depend on eligibility, clinical documentation, coverage requirements, information from a provider and a review that may require professional judgment. Each component can function well, and the overall process can remain slow because the difficult work often sits between systems.

Beginning in 2027, impacted payers face new CMS API requirements involving provider access, payer-to-payer exchange and prior authorization. The CMS Interoperability and Prior Authorization Final Rule requires impacted payers to implement certain FHIR APIs, with API compliance dates generally beginning January 1, 2027. The Prior Authorization API must support requests and responses and communicate whether a request is approved, denied with a specific reason or requires more information.

Those connections create a foundation for a more integrated operating environment. The information still has to enter a process that understands its context and knows where the work should go next. Software may be able to complete an action in one situation, while in another case it has to reach a person.

An AI agent could retrieve a clinical record, examine a coverage rule or recommend a next action. Its contribution depends on how well that intelligence fits into the complete operating process.

That is why I would be cautious about measuring AI progress by the number of agents deployed. Giving an agent access to five applications says very little about whether the underlying process improved. If people continue moving information manually and deciding where the work goes next, the organization has gained another intelligent capability without necessarily gaining much operating capacity.

Architecture should follow the work

I have spent years looking at automation through the way businesses operate. I tend to break an organization into its functions; understand the work each function performs and examine how information and responsibility move between them. The industry may change, although the discipline remains similar: start with the work.

For a healthcare CIO, that means understanding one complete process end to end. The CIO needs to know which information the process depends on, where it originates, who touches the work, which activities are repetitive and where judgment enters. From there, the architecture has to account for the system or person responsible for the next action.

These operating questions expose the difference between automating an isolated task and changing the way a complete workflow functions.

I organize that healthcare architecture around four layers: experience, intelligence, operations and secure data. Experience is where providers, payers, patients and care teams interact with the enterprise. Intelligence interprets information and can help understand what is happening or recommend an action. Operations carries that action through the organization by executing, routing, delegating or escalating the work. Secure data provides the records, rules and trusted information those functions depend on.

Their value comes from how the layers connect around the process being performed.

In prior authorization, clinical and coverage information has to be available when it is needed. Intelligence may help determine whether the information is complete and what it means in context. Operations then moves the work according to the result. Missing information can route the request back for completion, an automated action may allow the process to continue and a case requiring clinical judgment has to reach the appropriate professional with enough context to make that judgment.

Human involvement should be designed into the architecture from the beginning. Healthcare includes administrative work that automation can handle and decisions where professional responsibility stays explicit. Security, privacy, governance, compliance and monitoring belong in the same operating design. As agents participate in more of the work, the organization should be able to understand which information was used, what action occurred and where responsibility sits afterward.

This is where I think people-in-the-loop and agentic AI-in-the-loop become a useful operating model. The balance can vary across functions because the work itself varies.

Measure what changed

The measure I use for automation is what I call flex capacity. If one employee can handle 100 transactions and automation allows that role or team to handle 250, the organization has created additional capacity without adding people at the same rate. The business result is the increase in what the organization can do.

I have seen this in our own operating experience. Work that previously required roughly 500 people has been handled by approximately 100 to 150 people as automation changed the process. Coding, testing, data entry and data analysis are examples of activities where technology can reduce repetitive effort. Better visibility can also allow experienced people to oversee more work without the same administrative layers.

I would apply that thinking to healthcare workflows by establishing the operating baseline before adding agents. That baseline should show how many people touch the process, the volume the team can handle, the time spent moving information and the points where people intervene because the system cannot determine the next step. After automation, the same operating measures show whether capacity actually changed.

Greater volume from the same organization is one useful result. Reduced administrative work is another, particularly when clinicians and experienced professionals gain more time for work that requires their judgment. A large agent count provides much less information about the health of the operation.

The architecture has to support those outcomes across systems. An intelligent workflow may examine information, predict what could happen and recommend an action. Operations then carries the work to the appropriate destination. The path can involve software, a person or several systems depending on the process.

The 2027 CMS requirements give healthcare organizations a practical reason to improve important connections now. I would use that work to examine one complete operating process; from the information it depends on through the role intelligence can play and the points where professional judgment remains necessary. The resulting action has to reach the right destination, and the organization should be able to measure the change in operating capacity.

The earlier generation of healthcare AI taught me how closely intelligence depends on operating architecture. Today’s AI is considerably more capable, and healthcare interoperability continues to improve. CIOs have an opportunity to bring those developments together around the work itself.

The next advance in healthcare AI will be defined by how effectively the enterprise can put intelligence to work.

AI forces organizations to make consequential decisions during uncertainty. Roche chief digital and technology officer Wafaa Mamilli says this is the new leadership imperative. She believes success depends not on certainty but building the right foundations: trusted data, governance that enables speed, upskilling the employee base, and focusing on outcomes. At Roche, that approach drives applications ranging from kidney disease prediction and digital pathology, to new ways of discovering drugs and coaching people leaders. The challenge today, Mamilli says, is building organizations that can learn, adapt, and deliver value while the future unfolds, rather than predicting what’s next.

How is Roche using AI to drive growth and impact?

Our overarching mission is to prevent and cure diseases, and improve patient outcomes while reducing structural healthcare system costs. Data and AI are foundational enablers of this strategy. The high-quality data we’ve been assembling throughout our more than 125 years are part of this foundation.

We’re using data and AI to quickly make sense of a massive amount of information —  including genomic sequences, medical images, clinical records, and complex biology — and turn it into real medical breakthroughs at every stage of the patient journey.

We understand AI is a human and change management challenge, not a technical one. That’s why we prepared a holistic approach, which includes what we call everyday AI, to upskill our employees and embed AI in everyday work; reshaping with AI, which is reimagining our end-to-end processes; and our big ideas, which are large transformational investments.

Can you detail some real examples that bring this strategy to life?

We’re building AI tools across the entire value chain. Our enterprise platform, Galileo, includes internal generative AI tools, search engines, and compliance-safe sandbox environments.

We also have custom AI-powered coaching tools to help managers prepare for staff conversations. In our diagnostics business, we implemented an AI algorithm called Kidney Klinrisk to help physicians predict the risk of chronic kidney disease progression before irreversible damage.

Plus, we use AI image analysis in our digital pathology to assist pathologists in identifying cancer cells faster, and in R&D, we have AI-powered simulations to model clinical trial scenarios.

In manufacturing, we’ve deployed digital twins of our production lines, reducing development cycles by two to four times, and cutting technology transfer times by 50%. In drug discovery, we allow scientists to model over 100 experiments, compared to five in a physical plant.

But none of these examples imply that we’ve figured it out. We need to be confident and humble at the same time, but we’re still working through this just like anyone else.

How do you balance AI innovation and risk?

We develop life-saving medicines and vital diagnostics that directly impact human health, so we’ve always balanced speed and safety with the right governance framework. To position governance as a velocity enabler, we established a dedicated, horizontal framework focused on AI risk and ethics, and an AI strategy and governance function to ensure we leverage models responsibly.

But as important, I personally engage in understanding what’s at stake, because everything is changing overnight. That attention to detail is critical, because our job is to align, comply, and govern while allowing speed. We act with speed, assess, experiment, and then say yes or no to scaling the innovation. That’s how we lead this from the executive team down.

What changes have you made to your IT organization for this new wave of AI innovation?

I renamed my organization Roche Digital Technology, RDT, to position us as powering the business with digital tech and AI in a federated model. I created dedicated AI and data teams, which report to two new roles in RDT, chief AI officer and chief data officer. I intentionally made these two different roles because we need our data structured and ready for what we want to do with AI. There’s no AI without data, and we have foundational work to do there.

Think of the AI team as what the market calls forward deployed engineers — they work with all the other teams to re-engineer processes and deploy AI, while feeding the AI platform and tools teams so the platforms evolve to match the enterprise and marketplace. Enterprise architecture and portfolio governance are capabilities teams that support the full organization.

Describe the profile of the CAIO, as you’ve defined it on your team.

The AI officer is business outcomes driven and passionate, as well as tech savvy with high learning agility. This leader needs an entrepreneurial, risk-taking mindset, an ability to deliver, and deep process knowledge.

Remember business process re-engineering? There are some similarities here. We need people who can think through the workflow, not just what’s implemented in the systems but the actual workflow, what people do offline between systems, and think work and workflow, which we’ve never fully figured out. We need people who can step back, look end-to-end, and, with the knowledge of AI, reimagine what a process can look like now. So business process savviness is critical.

What advice do you have for tomorrow’s CIOs?

You have to be both a business and a technology leader. I call that a business minded technologist since business fluency is table stakes. You need to understand strategy, customers, and the economics of how the enterprise creates value. But particularly, in this moment, deep technology acumen matters enormously. The CIO has to understand not only what technology can do today, but what’s becoming possible tomorrow, and translate that into choices for the enterprise.

This is why I believe this is the most exciting time to be CIO. Technology and now AI is no longer enabling the business. It’s reshaping how companies operate, how decisions get made, how people work and increasingly how value is created.

So my advice for tomorrow’s CIOs is don’t be passive or apologetic about being a technologist. If there isn’t a chair for you at a table, bring your own by bringing an enterprise perspective and insights others can’t.

Few roles have the same vantage point across the organization. Use it. Connect the dots across functions. Understand the technology deeply enough to challenge assumptions. Bring confidence and humility, have the courage to make choices, and recognize that the hardest part of transformation is rarely the technology itself but bringing people with you.

The lobby is loud before the appointment begins. Dogs are barking, a phone is ringing and someone at the front desk is checking out with discharge papers in hand. A worried client is called into an exam room with a yellow Lab whose tail gives one soft thump against the floor, even though the dog does not really want to stand.

Inside the room, the client sits on the bench and hands the dog over for digital X-rays, hoping for the best while preparing for the worst. The veterinarian gently takes the leash, gives the dog a calm pat on the head and tries to project both confidence and compassion. The client speaks quickly, then more softly, trying to explain two days of changes: not eating, not wanting to jump, a strange cry when the dog was picked up, maybe a limp.

A technician is typing into the practice software on a laptop at the counter, trying to capture the history while keeping the visit moving. The team is listening, documenting, reassuring and preparing for the next step while the phone keeps ringing beyond the exam room door. This is the ordinary pressure of veterinary medicine: skilled people doing their best in a noisy, time-sensitive environment.

That is exactly where AI tools can start to look appealing. A tool that drafts the note, summarizes the history, organizes records, flags an image or generates follow-up instructions may feel less like futuristic technology and more like relief. Veterinary publications are already discussing AI applications in practice, including support for SOAP notes and workflow tools. In a profession stretched thin, a promise of efficiency is hard to ignore.

That is also why compassion is not a control. A caring team can still rely on a flawed tool, and a well-meaning employee can still paste an inaccurate AI-generated note into a record. Good intentions matter deeply in veterinary medicine, but they do not validate a vendor claim, protect sensitive information or define when an AI system should and should not be used.

Veterinary practice is the example, but not the only audience. The same pattern appears across smaller clinical, professional service and high-trust environments where AI tools are easy to adopt but formal governance may be light. For CIOs and technology leaders, the lesson is broader: AI governance cannot stop at enterprise frameworks if adoption occurs in environments where operational controls may be informal, inconsistent or missing.

AI will arrive as a convenience before it looks like a risk

AI adoption in veterinary medicine may not begin with dramatic clinical decision-making. It may begin with the ordinary, overburdened parts of practice: documentation, client communication, scheduling, reminders, inventory, imaging support and practice management. These are not glamorous areas, but they are exactly where small improvements can feel meaningful to a busy team.

That ordinary entry point is part of the risk. When a tool is framed as administrative support, practices may not treat it as governance-relevant. A note generator may seem like a convenience until it introduces an error into the medical record. A client communication tool may seem harmless until it gives confusing advice after surgery. An imaging support tool may seem like an extra set of eyes until someone begins relying on it more than intended.

The lesson for practice leaders is simple: AI is not risky only when it diagnoses. It becomes risky when it quietly shapes documentation, communication, data handling, workflow and accountability. Once AI influences what is recorded, what is sent to a client or what gets escalated, it is no longer just an efficiency tool. It becomes part of the operating environment.

That is why ethical and legal discussions about AI in veterinary medicine matter, but they are only the starting point. The harder work is translating those concerns into daily practice: which tools are approved, who may use them, what information may be entered and what outputs require review.

Good intent does not assign accountability

Veterinary teams are used to carrying emotional weight. They comfort clients, handle difficult decisions, work through emergencies and continue moving from room to room. That culture of care is one of the profession’s strengths, but it can also make the risks of technology harder to see clearly. When everyone is trying to help, it is easy to assume the tool is simply helping too.

AI does not remove accountability from the practice. If an AI tool drafts a record, the practice is still responsible for the record. If a tool summarizes a client conversation, the practice is still responsible for what is retained and acted upon. If a system flags a possible finding on an image, the veterinarian is still responsible for interpreting the patient in context. If a chatbot responds to a client, the practice still owns the boundaries of that communication.

Those boundaries should be explicit before the tool is used. Staff should know whether AI-generated notes are drafts, who must review them, whether AI-assisted content can be copied into the medical record, and what kinds of client questions require a human handoff. Without those rules, accountability becomes implied and implied accountability tends to fail under pressure.

Someone also has to know which AI tools are approved, what data they collect, how they are used and how concerns are reported. That person does not need the title of chief AI officer. In many practices, it may be the owner, medical director, practice manager or another designated leader. What matters is that oversight is named, not assumed.

Vendor claims deserve the same clarity. Before adopting an AI tool, leaders should ask what the tool does, what it does not do, how it was tested and what data it uses. In human medicine, the FDA maintains a public list of AI-enabled medical devices authorized for marketing, showing how validation and transparency are treated in adjacent clinical technology environments. Veterinary tools may not always follow the same regulatory pathway. Still, the governance question remains: how does the practice know the tool is appropriate for the way the staff intends to use it?

Lightweight controls can protect trust

The answer is not to bury veterinary practices under enterprise bureaucracy. A small clinic does not need the same structure as a hospital system or a large corporation. What it needs is a practical operating model that fits the practice’s size, risk and reality.

That model can start with a short list of approved AI tools and a rule that staff should not use unapproved tools for client, patient or practice information. It should define what AI may be used for, what it may not be used for and which outputs require review before they are entered into the medical record or reach a client. It should also specify who is responsible for oversight and how staff should report AI-related concerns.

Data rules should come before convenience. A client may say something personal during an appointment, and a record may include financial details, legal concerns, rescue history, breeding information, animal welfare issues or emotional context that was never meant to leave the practice environment. Veterinary practices may not operate under the same rules as human healthcare systems, but their data still matters.

Reporting matters because AI errors may start small. A note may contain a detail that was never said, or a summary may leave out the owner’s observation that changes the clinical picture. If staff do not know how to raise those concerns, the practice may normalize small failures until a larger one breaks trust.

AI may become useful in veterinary medicine, especially if it reduces administrative burden and gives skilled professionals more time for the work only they can do. The profession is already exploring AI’s potential, challenges and future direction across practice types and species, and technology that genuinely helps veterinary teams should not be dismissed out of fear.

But adoption is not the same as readiness. A tool can be helpful and still require oversight. A vendor can be innovative and still need validation. A caring team can use AI with the best intentions and still create risk if no one has defined how the tool should be used.

That is why compassion is not a control. It is not a cynical statement; it is a protective one. Compassion is essential to veterinary medicine, but compassion alone cannot review an AI output, secure client information, evaluate a vendor or decide when convenience has outrun judgment.

The practices that benefit most from AI will be those that protect trust as they adapt. AI can support the work, but it should not quietly redefine it. Before AI becomes another voice in the exam room, veterinary practices need to decide how to govern it.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

One of the biggest challenges companies like Aetna face every year is an annual HEDIS review of its records to identify gaps in care. For large national payors, the scale of the challenge is immense. So Aetna has deployed a gen AI-driven document intelligence platform that has reduced the need for manual review by 65%.

“We have a large group of amazing trained medical coders who do this every day,” says Nathan Frank, chief digital and technology officer at Aetna. “This is about making it easier for them by speeding up the process. Something that might have taken weeks or months we can now do in days.”

The Healthcare Effectiveness Data and Information Set (HEDIS) is a range of performance measures for the managed care industry. Developed and maintained by the nonprofit National Committee for Quality Assurance (NCQA), the first version of HEDIS was released in 1991.

Under the HEDIS measures, large managed care providers like Aetna review more than 10 million medical records annually to identify gaps in care. These gaps are missed or overdue preventative care or chronic disease management tests including missed cancer screenings, blood sugar tests for diabetics, eye exams, and immunizations. Closing these gaps improves patient outcomes, and health plans are measured in how well they perform. But processing medical records is no easy task.

“We’re talking about medical charts that have white space filled with handwritten notes,” Frank explains. It’s not just structured data, it’s lots of physical clinical documentation.”

Adding up the numbers

Frank says industry benchmarks for large providers indicate an annual review process that requires about 50,000 work weeks, equivalent to nearly 1,000 dedicated full-time employees. It would take a team of 50 reviewers more than 20 years to complete a single annual review using fully manual processes.

Enter AI Medical Chart Review, a platform developed by Aetna that leverages cloud services and gen AI to automatically extract clinically relevant data from records, and prioritize records based on the likelihood of measure closure and evidence strength.

“Large language models and gen AI give us the ability to train a model to decipher the charts, identify the high value codes, and build correlations,” Frank says.

In the space of about six months, Frank’s team ideated the platform, and designed and trained a PoC that was able to process millions of records in just two weeks. As a result, AI Medical Chart Review has earned Aetna a CIO 100 Award in IT innovation.

“Now we’ve gone through 14 million documents,” Franks says. “We’re seeing a reduction of manual effort, which is now being transitioned into other areas like quality control and making sure the automated chart review is working as expected.”

Behind the curtain

Using gen AI, the platform automatically ingests and analyzes unstructured medical records and clinical documents. And as part of that process, it identifies and extracts clinically relevant information for specific HEDIS measures like diagnosis codes, medication records, lab results, and visit documentation. With this data, the platform generates a prioritized set of records based on the likelihood of measure closure and strength of clinical evidence, which is then passed to human employees for review and validation.

Frank says the platform has increased gap closure rates (leading to improved Star Ratings and higher reimbursement), streamlined workflows, and enabled teams once dedicated to manual record review to shift focus to higher-value activities.

Frank says much of the speed and success in building the platform comes down to a shift in the way it approached the design and build process. Rather than exhaustively writing specifications and requirements, Aetna created a team that included engineers and subject matter experts who worked together to build out capabilities iteratively.

“It allowed us to move much faster, and having a business subject matter expert sitting in the same virtual or physical room with us got us a much better outcome,” Frank says. “The product model, our cloud compute model, and our AI governance model allow for quick reviews to make sure we’re using AI responsibly with the right guardrails. It’s increased the speed to get from product launch to go-live.”

He adds that small teams that don’t have to deal with a lot of bureaucracy are key to moving quickly.

“You need to design with security, compliance, and a responsible use of AI as core principles from day one,” he says. “Everything we do from a new build standpoint starts with thinking about how we make it cloud native, how we build with the right elasticity and speed, and how we optimize the cost.”

The most important element of all, he says, is a good relationship with your subject matter experts.

“You can have a great product manager and engineer, but you really need that business subject matter expert who’s excited about it, and who has a passion for transforming the process,” Frank says. “Once you put those three together, you’ll see amazing things like this happen all the time.”

Medical technology companies have always been in the business of trust, and Abbott has been building it with AI for over 10 years. Long before gen AI entered the enterprise conversation, Abbott was using algorithmic AI to help diabetics manage their glucose, and imaging AI to guide surgeons in real time. Here, Sabina Ewing, Abbott’s CIO, explains how a principled approach to AI governance, deep cross-functional partnerships, and a commitment to demonstrating results from within IT have kept them ahead of the curve, and its mission intact.

How is Abbott using AI to achieve its mission and growth strategy?

As a medical technology company, Abbott’s mission is to help people live life to the fullest. For over a decade, we’ve been using AI to deliver on that mission, but whether it’s AI or any other technology, we’re intentional about how it ties to our mission.

Trust is earned in drops and lost in buckets. To ensure we maintain trust with our customers and employees, we’re guided by principles of fairness, safety, quality, and transparency. With these and our mission as our guide, we’re in command of the table we set for ourselves.

How have you been in the AI business for so long?

For decades, we’ve provided FreeStyle Libre, a glucose monitoring sensor built on algorithmic AI, that delivers continuous glucose readings to diabetics, and in some instances, connects to insulin pump applications.

In late 2025, we developed Libre Assist, which leverages generative AI to let FreeStyle Libre users take pictures of their food and receive guidance on the impact of that meal on their glucose levels, including when to eat what, because sequence affects how the body processes glucose.

In our medical devices business, Ultreon, launched in 2021, uses imaging AI to guide optimal stent placement during cardiovascular procedures, supporting the physician’s decision-making in real time.

So whether it’s algorithmic, generative, or agentic, we’re intentional about matching the capability to the specific therapeutic problem.

When technologies evolve, your mission doesn’t change. But how has the CIO evolved during this AI boom?

Today’s CIO must have the strengths of conviction, credibility, and communication. You need technical expertise and to surface data to have the right discussions. You also need to be brilliant on the fundamentals and clear about the strategy, and then execute against it. If I tell the business it can use AI to drive outcomes, then I need to demonstrate it in IT. This is why I’ve committed two commas of results in IT from new AI operational capabilities.

How can CIOs influence their company’s investment in AI?

Working with senior leaders in HR and finance ensures we’re educating the organization and securing necessary investment, and then maintaining financial discipline where investments occur. We hold to that discipline and we’re deliberate about how we deploy the resources of the organization to measurably have impact. We look for high-impact opportunities where new technology delivers results even as it evolves.

We established an executive steering committee on generative AI, and senior leaders are engaged in how we deploy capital. We’re not going out with a thousand flowers blooming.

We also have traditional financial measures we apply to AI investments. And we know you need to be able to identify quantifiable outcomes and then measure them. Those conversations happen in partnership with all senior leaders, especially those business leaders requesting specific capabilities.

As the CIO, you need to have strong relationships with all other parts of the company. I don’t need to be in the spotlight, but you need those relationships in order to lead and effect change. I led a program that helped educate our top leaders on AI foundations, and we’re all working together on the talent side, too. We’ve embedded AI into our talent processes, and we have a continuous cycle of enterprise education through the ranks, both in person and virtual, to ensure our people are ready to use the latest tools and technology. If you want to do something sustainable, you can’t do it by yourself.

What’s your message to your technology team?

What I tell our team is no one is better positioned to lead the organization through this transformative era than its own technical experts. That means we lean into our expertise and AI-first mindset.

Years ago, we crystallized our vision for Abbott IT by unleashing the power of technology and our people in service of Abbott’s purpose. That’s the constant reminder. Our role isn’t to deploy tech but to unlock what technology and people can do together, in service of the mission.

I have asked the team to be bold, bring their best, and pursue excellence. Our strategic pillars are modernization, and protecting Abbott in both enterprise and product cybersecurity, digitization, and advanced analytics. That’s always been part of our mandate.

But what does an AI-first mindset look like? I asked our executive assistants how they, as a community, think about using AI to radically expand what they can do with it as a companion to their work, but not a replacement for it. The models that exist today can’t be a great executive assistant. Models don’t have the judgment, institutional knowledge, or nuanced reasoning required to prioritize work and navigate unspoken rules. That expertise is irreplaceable. Our question is how to augment it.