Vikram Anand explains why healthcare data management remains difficult and how standardization, MDM, human review and AI can improve patient care.
During a recent conversation on The Executive Outlook, Vikram Anand shared a lesson shaped by more than fifteen years of working across healthcare, data, and artificial intelligence.
Technology is rarely the hardest part.
Early in his career, he worked on computer aided diagnosis systems that analyzed medical images and helped identify findings such as lung nodules and colon polyps. At the time, teams did not have foundation models or readymade tools available today. Much of the intelligence had to be built from the ground up.
The experience proved that AI could support faster analysis and help clinicians process complex information. It also exposed a deeper problem. The performance of any AI system depended on the quality, structure, and availability of the data beneath it.
That insight has followed Vikram throughout his work in population health, voice-enabling healthcare, predictive analytics, and transplant care.
Healthcare can build powerful AI, but it cannot create reliable patient outcomes if the underlying information is fragmented, delayed, incorrectly matched, or difficult to access.
Healthcare has made meaningful progress in digitization. Electronic health records have replaced much of the paper-based process. Standards such as FHIR have improved the way information can be exchanged. Health information exchanges have been created to move data across hospitals and care networks.
Yet daily reality remains complicated.
A patient may complete a laboratory test at one location and receive treatment at another. The test result may exist digitally, but the clinician who needs it may not receive it at the right time or in a format that is easy to use.
Medical images may still travel on CDs or DVDs. Documents may still be sent by fax. Important details may remain buried inside electronic records or unstructured text.
This is one of the clearest healthcare data management examples. Digitizing information does not automatically make it accessible, interoperable, or useful.
He sees the challenge most clearly in transplant care. Transplant centers often operate as central hubs, while patients live in other cities or remote areas and complete tests at different hospitals and laboratories.
Before care can progress, coordinators must confirm that the right tests were completed, locate the results, verify the dates, understand the clinical meaning, and connect every piece of information to the correct patient.
When healthcare data interoperability fails, the burden moves to people.
Coordinators spend hours searching, calling, checking, and reconciling records.
This is not simply an infrastructure issue. It can affect how quickly a patient receives care. The real question is whether the right information can reach the right clinician at the right moment.
He describes the foundation beneath healthcare AI as the unglamorous plumbing.
It includes data movement, integration, storage, governance, security, standardization, identity matching, and regulatory compliance. These activities rarely receive the same attention as a new AI model or clinical application, but every intelligent workflow depends on them.
Business leaders may see a powerful AI tool and want to deploy it quickly. Technology teams see the work required to make that tool safe and useful.
The difference creates a mismatch between business expectations and technical reality.
In many industries, an imperfect AI response may create inconvenience. In healthcare, incorrect information may influence a diagnosis, treatment decision, or the timing of patient care.
That is why healthcare organizations cannot simply add AI on top of fragmented data and expect trustworthy results.
The first executive question should not be whether the model is advanced enough. It should be whether the data is complete, standardized, traceable, and suitable for the decision the model will support.
Healthcare master data management becomes essential when organizations try to create one trusted view of a patient.
The same person may appear differently across systems. A name may contain a missing space, a spelling variation, a different order, or an incomplete field. Even a small inconsistency can prevent systems from recognizing that two records refer to the same individual.
Clinical information adds another layer of complexity.
Different systems may use different medical coding standards. The same condition may be represented in several ways. One source may describe diabetes through one terminology, while another uses a different code or wording.
Master data management in healthcare must reconcile these differences without attaching information to the wrong person or changing the clinical meaning.
He compares the volume of information associated with a transplant patient to reading a book like Moby Dick. The record may include laboratory results, imaging, diagnoses, medications, referrals, evaluations, and information from several facilities across different periods.
The objective is to create a golden patient record that brings together the most reliable version of this information. This is a continuous process of matching, validating, structuring, and reviewing data because one error can travel through every downstream workflow.
A laboratory value of 7.5 cannot be 75. The correct unit must remain attached. The test date must be accurate because clinical results may have a limited period of relevance. The information must belong to the correct patient.
Patient data management therefore requires both scale and precision.
AI can make healthcare master data management faster.
It can help match records, extract information from unstructured documents, summarize long patient histories, identify clinical values, and transform text into structured data.
These capabilities are valuable when the volume is too large for people to process manually. But he is clear that someone must review the output.
Healthcare information changes continuously. New documents arrive, coding practices evolve, systems change, and models encounter situations they were not originally designed to handle.
A solution that worked correctly at launch may not remain reliable without monitoring and improvement.
This is why he describes the future of healthcare data work as a shift toward stronger review. Professionals may spend less time manually extracting every value and more time confirming that AI has interpreted, matched, and structured the information correctly.
The goal is not to remove people from the process. It is to use AI for speed while preserving human judgment where accuracy matters most.
AI is already creating value in healthcare.
He has worked on computer aided diagnosis systems that analyze medical images and support the detection of findings such as lung nodules and colon polyps. He also discussed how some screening workflows have become increasingly machine-driven, with humans reviewing uncertain or exceptional cases.
These systems can reduce analysis time and support clinical teams at scale, but their success depends on healthcare data standardization.
Medical images must be stored in recognized formats such as DICOM. Clinical values must include the right units. Dates, identifiers, and terminology must remain consistent. Data must be available in a format the system can interpret reliably.
When every organization store information differently, AI becomes harder to deploy and maintain.
He summarizes the principle simply. If poor quality information goes into AI, poor quality output will come back.
For executives, this is a question of ownership. Before approving an AI initiative, leaders must know who is responsible for the quality of the data, how errors will be detected, and what happens when the system is uncertain.
He sees significant potential for agentic AI in healthcare, particularly in administrative and coordination work.
Care teams spend considerable time calling patients, checking whether laboratory tests were completed, sending medication reminders, following up on appointments, and locating information from external facilities.
Many of these tasks are necessary, but they also consume time that clinicians and coordinators could spend on more complex patient needs.
AI agents could support these activities by contacting patients, answering routine questions, providing reminders, checking whether an action was completed, and helping retrieve the required information.
He connects this possibility to his earlier work with Alexa Health, where voice technology was used to support elderly and vulnerable patients with daily activities and medication related questions.
The value of agentic AI in healthcare is not limited to replacing a phone call. A patient can receive a reminder at the right moment; a coordinator can see whether a task was completed, and missing information can be identified earlier.
However, greater automation also creates greater responsibility.
The agent must use the correct patient's record, follow privacy requirements, understand its limits, and transfer the interaction to a person when necessary.
Trust is one of the most important conditions for healthcare AI adoption.
He does not suggest giving AI complete control from the beginning. He recommends introducing it alongside existing human processes.
AI may handle routine interactions while a person remains available for exceptions, concerns, or situations where the patient prefers human support.
Choice matters.
Some patients will quickly accept automated systems. Others will remain uncomfortable, even when technology performs well. Healthcare organizations must respect that difference.
People should be able to ask for a human.
Trust grows when patients and clinicians see consistent performance, understand what the system is doing and know that human support remains available.
The provenance of the data must be clear. The information must be correctly structured. Regulations such as HIPAA must be followed. The system must be tested in the workflow where it will actually operate.
Trust is not created by announcing that an AI tool is accurate. It is earned through reliable performance over time.
Vikram’s message to healthcare executives is direct.
Start with the data.
Understand what information the organization has. Know where it comes from, how it is stored, how it is standardized and whether it is available to the people and systems that need it.
Use AI to accelerate extraction, matching, summarization and structure. But build the review capability required to confirm that the output is correct.
Do not treat healthcare data management as an isolated technology programme. It is part of patient care.
Every platform, integration, governance rule, and master data process should serve the same outcome. The right information must reach the right clinician at the right time so the patient can receive the right care.
That is the foundation healthcare AI needs. Organizations that build it can move beyond isolated pilots, while those that ignore it may keep investing in advanced tools without resolving fragmented records, delayed decisions, and manual work.
Vikram’s final insight captures the change already taking place.
Healthcare professionals are becoming better reviewers, not simply better coders or data extractors.
The leaders who recognize that shift will build systems that combine machine speed with human responsibility.
In healthcare, that combination is not optional. It is what makes innovation safe enough to matter.
Want to hear more conversations with leaders building data foundations that make AI work in the real world? Explore more on The Executive Outlook.
Listen in for business strategies and interviews with business experts to help you build a business and a life that you love.
About Us
