Artificial intelligence in healthcare is no longer limited to chatbots or back-end analytics. It is now moving into something far more sensitive, real-time payment decisions. That means AI systems are helping insurers decide whether a medical claim should be cleared even before money moves. For an industry that handles billions of claims and operates under tight regulation, this is a big shift. To understand what is really changing on the ground, I spoke to Subha Vaidyanathan, VP at EXL, about how AI is being used in healthcare payments, how accuracy is maintained, and where humans still play a critical role.
Below is the full interaction.
For anyone who may not be familiar with EXL, can you give us a quick overview of the company and the kind of work you do across industries?
Subha Vaidyanathan: EXL is a NASDAQ-listed company and we’ve been here for 27 years. And we have close to 63,000 employees and we actually engage in multiple verticals. We provide IT services solutions across banking, healthcare, insurance, and different verticals.
Also read: Pichai and Hassabis: India uniquely positioned for AI leadership
A lot of companies talk about AI in healthcare, but you’re saying it’s now making real-time payment decisions. What changed recently that made this possible at scale?
Subha Vaidyanathan: One of the problems we had in healthcare earlier was the data silos that we had. Fragmented data was actually creating a lot of leakage, dollar leakage, especially in the system. So we right now have platforms which can get this particular data, do the transformation of the data, structured and unstructured data in the form of claims and in the form of medical records, and have the data processing much faster. So using our platform, 60% of the velocity is actually increased, making decisions more near real-time. And we are creating this impact in the healthcare industry, where we are moving from a pay-and-then-pursue model to a more shift-left and then do the analysis and then do the payment. So this shift-left in the industry is being very favoured by our clients.
But this is happening globally, right? What about India’s plans?
Subha Vaidyanathan: Yes, we are evaluating the Indian market right now. We are more focused on creating AI talent, but we are observing the market and evaluating it, especially in healthcare.
There’s a growing push to stop payment errors before money even moves instead of fixing them later. How does AI actually help catch these issues early, and where do humans still come into the process?
Subha Vaidyanathan: If you look at it, when a patient and a physician encounter happens, it is digitised and it is called a structured claim. And then we have the episode, which the doctor is recording, which is called the medical record. So both of these are analysed and we detect all of the anomalies which are happening, and this is being done in real time. This is shifted left, and we have AI solutions which are helping look at the medical record, being able to index the medical record, divide it into multiple sections, and extract exact information which is required. After analysing it, we do the auditing of that particular claim and the medical record to get a human-in-the-loop closure because it is, after all, healthcare of people and we have to be very, very cautious about it. So that’s where our human-in-the-loop auditors work. They finalise the payment, which is done for the claim, and that’s why we have the human-in-the-loop scenario.
What accuracy are we looking at?
Subha Vaidyanathan: It’s more than 99% accuracy that we have to adhere to. We are a very highly regulated industry. So that’s the percentage we look at.
When AI is handling billions of claims, trust becomes a big factor. How do you make sure the system stays accurate and compliant, especially in such a sensitive space like healthcare?
Subha Vaidyanathan: True. So we have a zero-trust policy in terms of giving access, but then validating and providing that particular access. And also, since we work in a highly regulated industry, we follow all of the norms which are there, all of the statutory norms in terms of high trust, HIPAA certification and SOC certification. Our environments are certified. So zero trust helps us achieve that.
Everyone talks about ‘human-in-the-loop’ AI, but what does that really look like day to day? Where do you think automation should stop and human judgement should take over?
Subha Vaidyanathan: So, this is an example for the analysis part of it. We process billions of claims and millions of medical records. So we use AI to do all of the grunt work. For example, extraction. It eliminates the manual labour for a person to look at that particular medical record and extract all of the data. That’s where we are employing AI. And then the decision-making is still with the humans, so that the accuracy is higher. That’s what we are doing.
Also read: India AI Impact Summit 2026: What is agentic commerce? Mastercard’s Nitendra Rajput explains