Kiran Veernapu has spent well over two decades building systems that extract millions in savings from healthcare operations. His track record spans aviation supply chains, enterprise software, and multi-hospital data platforms, which is work that produced roughly $17 million in documented cost reductions and a 30% drop in surgical billing errors. He’s also reviewed over 60 manuscripts for IEEE, Springer, and Elsevier, giving him a front-row view of where AI research meets operational reality.
We spoke with him about the gap between promising pilots and implementations that actually stick, why supply chain remains healthcare’s most undervalued AI opportunity, and how his peer review work shapes the way he evaluates whether an AI initiative is ready to deploy.
Kiran, you’ve spent 25 years working across healthcare, aviation, and enterprise technology. How has that cross-industry experience shaped the way you approach AI implementation in hospital systems?
My experience has evolved since the beginning. I’ve always had a great passion to be innovative and make the software solutions easy and simplified for a better user experience in any domain, from aviation to healthcare. The drive is to think out of the box and provide a solution using the latest cutting-edge technologies. This shaped up my career providing simplified solutions to industry problems.
Enterprise technology helped me learn the way systems can be designed to be scalable, to achieve high performance, and be very protective for data access. Aviation process automation helped me understand how critical it is to be accurate and provide data driven decisions.
Healthcare is very critical for patient safety and time sensitivity to make decisions in a timely manner. Data is always the key for decision making. One cannot make mistakes in providing clinical advice. AI helps in all of these areas to enhance the solutions to be efficient and effective where there is less possibility for error and gives scope for learning the data patterns and suggest the specific use case solution.
Your work at a large nonprofit health system has generated roughly $17 million in cost savings. Where do most health systems leave money on the table when it comes to data and AI infrastructure?
There are several ways healthcare organizations lose or spend money that can be evaluated for efficiency and saving millions of dollars with proper data analysis and evaluation. The high amount of cost run down happens with supplies and supply chain cost. Analyzing each segment of the supply chain and its impact on a patient cost like surgeries or procedures gives more insights for savings. Use of AI in understanding the gaps and automating the supply chain with AI solutions helps save 40% of the healthcare cost these days and also makes the healthcare system run efficiently.
Secondly, analyzing the operational cost of running healthcare is the key for reducing the cost of healthcare. The value comes when AI solutions are truly embedded into the solution. Some examples are automating the insurance claims, reducing the errors in preparing the claims
Third is the building of integrated systems to avoid fragmented systems and islands of data that give a variety misleading cost calculations. The result of this is a missing opportunity in inventory optimization, care coordination, and revenue cycle optimization.
The biggest opportunity in healthcare is making data actionable at the point of decision making.
What does it actually take to consolidate data across a multi-hospital system without disrupting clinical operations in the process?
The important step towards data consolidation is to identify the isolated data systems, planning for the common data model and governance framework that works well for multi hospital systems. Designing and mapping processes to align for a common framework is important.
The architecture needs to be non-disruptive, and an unified layer on top of the existing system. Architecture like Fast Healthcare Interoperability Resources (FHIR) using real time APIs for healthcare systems. This helps hospitals continue running as is and while the data is integrated in the background.
The clinical workflow needs to be preserved while enhancing the functionality of the healthcare system so that the clinicians do not need to learn all together a new system. This can slow down the care and the clinician’s adoption rate for the new system will go down.
The systems to be implemented in an incremental fashion by choosing one healthcare system first and with successful implementation, the solutions to be implemented at large on multi-hospital systems.
Supply chain optimization in healthcare doesn’t get the same attention as clinical AI, but you’ve done significant work there. What are people missing about that intersection?
There are several areas in supply chain management for hospitals. The important steps are the supply of healthcare material is critical for treatment of patients. Hospitals housing too much inventory or inventory being unavailable both are going to give adverse financial, and operational results, also they are patient safety concerns.
The use of AI in optimizing the supply chain in healthcare is the biggest cost saver in running health systems. Examples are implementing the twin bin Kan-ban system, which helps maintain the appropriate stock levels, and replenishing at the right time to keep the right stock levels. Making automated reorders with a data-driven approach. This saves cost on expired medical equipment, unaccessed inventory for several months, etc.
Using AI in reconciling the operative inventory, sometimes there are minute parts used in the surgery where clinicians may not be able to remember to code in billing or excess coding issues. This gives 30% of error reduction in operative inventory.
AI engines can constantly search for the cost optimization with suppliers to choose the right supplier with quality and cost efficiency. Bench marking supplier performance with the inventory and historical invoicing can help identify the savings to hospitals to be able to negotiate the product cost with suppliers.
AI engines help understand inventory cycles to be able to maintain appropriate stock levels with hospitals spanning across multiple cities and locations. It will be very effective to integrate and use AI solutions on optimizing supply chain operations.
You’ve reviewed manuscripts for IEEE, Elsevier, and Springer, and published 25 of your own papers. How does that research lens change the way you evaluate whether an AI initiative is actually ready to deploy?
AI initiatives are evolving and solutions are gaining confidence in the eyes of clinicians. There is an enormous amount of research being conducted with proven data samples. Every healthcare system and process can be benefitted by adopting AI solutions. The ultimate goal of AI is to assist in an efficient way where the human eye cannot reach or visualize. AI can help and not harm healthcare systems. The key is to train the models with large amounts of data in providing assistance.
In my experience, AI solutions must provide rigor, reproducibility, and real-world impact. Some inputs, such as whether the methodology is effective, are the results valid statistically, can this experiment be preproduced in another setting?
Explainable AI with accountability can help clinicians to trust the models and the outcome. AI initiative will fail if the model is weak and the results are not reproducible in real world application.
Most health systems have tried and stalled on AI projects. What separates the implementations that hold up from the ones that quietly get shelved?
When AI projects are started as mere technology improvement they tend to stall. While AI implementations that are reproducible in the real-world scenario are successful, these solutions are embedded into electronic health record solutions.
When AI solutions are compared to the human evaluated results and if they are close and realistic they are trusted by the clinicians, the rate of error can be reduced and improved over the period of usage.
Several AI projects fail when they have fragmented and isolated and inconsistent data. When systems are implemented on an integrated system AI provides best value.
