What we know is speed is critical. So any use case that you pick up or any implementation, you have to be able to make the return on that investment be visible in a very short period of time. And these days, that short period of time is as little as 30 days. And so, you know, you really need to move with high velocity.
You're listening to Make It Real, brought to you by EXL. I'm your host, Shirley McBTH, and on this show, we're exploring how artificial intelligence is reshaping workflows, industries, and the way real work gets done. And yes, we're going to make it real.
Everyone's talking about AI and wants to use it in their organization, but the reality is that many leaders are struggling with how to use AI effectively in their organization and get the real benefits from it. So joining me today is Rohit Kapoor, the CEO and chairman of EXL, a global leader in data and AI. And Rohit really brings a unique perspective to this conversation through the work that EXL is doing to empower companies that are really in complex industries to transform their businesses with AI. So we're going to dive right in and talk about why that conversation around AI is really important and critical right now. And then we'll also tackle how to overcome those barriers that often hold organizations back from realizing that great potential that we see from AI. So welcome Rohit and thanks for joining us today.
Thanks Shirley. It's exciting to be here on the first podcast by EXL on this topic.
Excellent. Well, thank you so much. So Rohit, you talk to executives all the time about AI adoption. And you're out there talking to many leaders and you know there's that promise of AI, everybody's very excited and then we see there's studies out there that talk about there's a definite gap between that hope and the potential of AI and what is being realized by many organizations today. Can you talk about what's causing that gap?
Sure. So actually it's very simple. The basic promise on using and leveraging AI is that the return on investment is going to be phenomenal. And I think what organizations are finding is it's actually very hard to deploy AI and getting that return on investment is actually quite elusive if you don't implement it the correct way around. So many use cases that clients have tried to implement AI for just does not provide the return. And I think you can characterize this and you can place them into three different buckets as to why does this happen. Reason number one is many of the processes where AI is being implemented, it's being implemented as a point solution trying to get productivity benefit for a particular task. And therefore what happens is while each task provides the productivity benefit and the promise is being realized at the task level, you really don't get the benefit at the process level and there is no real cost savings that drops down to the bottom line because the same person is doing part of their job in a much more efficient way, but then they're also doing the rest of their regular job in the same way as they did before. A second thing is in many situations it's not the application of AI across the enterprise. It's only applied to a small subset of processes and operations where the AI is being applied. And that means that you know you're not really getting scale benefits and you're not really getting the real power of AI coming through in terms of that return. And then finally many times the choice of the use case is such that the benefit that can be provided by the use of AI is very very limited. So it typically ends up being that the return on investment is actually quite low and that's why clients don't feel like embracing AI as aggressively as they ought to be. The trick is pick the right use case. Make sure that when you implement AI, you're not only implementing AI, but you're transforming the process. So that means you're going to handle the process and you're going to work on that process differently. And by reconfiguring the process, you will be able to shrink down the workforce and be able to get a realizable benefit associated with that. And then lastly, work on those types of use cases that can be leveraged across the enterprise. Whether that's a horizontal application of AI or whether that's a vertical application of AI, but it needs to extend itself either horizontally or vertically and provide the necessary benefit to the client.
You said so many great things there. So I want to unpack that a little bit. So, I heard you say when you're thinking at the task level, you're not getting that full benefit that you really need to think bigger and across horizontally and deeper for some of the broader use cases. And I also heard you say pick the right use case is critical to this. So, let's unpack that a little bit. How do you advise executives and CEOs to think about what's that right use case? And it seems daunting, right? So, something that's horizontal and huge. How would you approach that?
So the choice of the use case is absolutely critical and I think it'll get driven by a couple of fundamental factors. We should pick a use case which is a large problem and an area where there can be a significant amount of efficiency or effectiveness or an ability to have a significant amount of growth associated with it. So it's got to be something which is meaningful. Second, the use case has to be such which means it is ready for the application of AI and that involves having data being available to apply the AI to it. Many times we find our clients tell us I'd love to be able to kind of deploy AI in a particular use case but their data is sitting across multiple silos. It's not accessible and you simply cannot apply AI in those situations. You also have to pick and choose those use cases where it will be easy to demonstrate the value of the application of AI very very easily and quickly. We have a diagnostic tool that we use to help our clients figure out where should we start and deliver the benefit of the use of AI to them and then once that credibility has been established then we can actually take that forward in the organization and apply that across multiple use cases.
That makes a lot of sense. I want to probe deeper on data. You talked about the importance of data and you know we know that data fuels AI but often it's very challenging to get that data house in order if if you will. Can you talk a little bit more about data and why that's important and what are some of the areas in particular with things like unstructured data that are critical to get in order as you move towards AI.
Yeah. So you know if you think about the application of AI there are two things which are critical. There's data and then there's the algorithm or the LLM. And 90% of the problem and the effort relies in getting your data estate in order. Most organizations have grown through acquisitions. They've got legacy infrastructure and technology platform sitting there and therefore the data sits in multiple silos and it's not interconnected and it doesn't talk to each other. Further, the data is not clean and it's not standardized. So you might have fields and records on one platform being used in one particular way and on another platform in another way. You also have the unstructured data issue that you spoke about. Almost 85% of all data that exists in an organization is sitting there as unstructured data. Whether that be in the form of handwritten notes or that be in the form of call logs or that be in the form of voice interactions that you've had with your customers. It's all sitting there. So what we find is only 30% of organizations say that they have data available in a usable format. And so the very first task is to get that data estate in order. And what we have found is that one of the best use cases of applying AI for any organization is to use AI to get their data estate in order so that data is ready for AI. And so it's literally about applying AI to get your data ready for AI. And we've developed a suite of agentic solutions which we call as EXL data.ai. These are about 65 different agents that we've put together and that allows us for the right amount of data discovery, the ability to have data lineage be established, have data governance, have data cleansing, data labeling. So it's a very comprehensive capability that we've created that brings down this task of having your data estate be AI ready down from a few months and bring it down and bring it down to a few weeks.
That's incredible. Well, when you think about using AI to solve your AI challenge, that's very exciting. And the time shrinkage of how you're able to do that, that's phenomenal. Let's talk a little bit about some of these workflows. You've been talking about workflows and sort of these big horizontal areas that can be transformed. When you think of something like a big healthcare organization or an insurance are these very complex a bank, can you give a few examples of the types of use cases that can be transformed using AI?
Sure. So, I'll give you two examples. The first one is what we did for an insurance carrier. They were receiving a number of emails with attachments associated with these emails and their underwriters had to literally open thousands and thousands of emails, open the attachments, reach out for the relevant information that they needed to make an underwriting decision and then price a policy and then go back to their customers and offer a price associated with that insurance policy. Well, we took that entire journey and now we are using Agentic AI to quickly read through all of these emails, open the attachments, pick up the relevant fields and the necessary input variables and be able to reduce the cycle time from a week to a few hours. So what that does is it allows the insurance carrier to go back to the broker with a quote and a much faster response time. And you know this that in today's day and age, if you ask for a solution or answer to a question, if you get that response time quickly within a couple of hours, your likelihood of making a decision is so much higher. So what we've seen is that the quote to conversion ratio has gone up by 7%. Which is huge for an insurance carrier and it's enabled them to process a volume which is three times the normal volume that they would handle in any other normal operating environment. So this kind of a deployment has been really enormous for this insurance carrier. Another example would be we work with a utility company in the UK which is one of the largest utility companies in Great Britain and what they have done is you know worked with EXL to embed Agentic AI across their workflow across all of their customer journeys. So today the onboarding of a new customer that's done leveraging AI. The entire process which is from meter reading to the cash payment that's done you know using AI. When a customer moves location and they want to close and shut down their service and want to move to a new location all of that is handled by AI. So today we've got almost about 35% of all of their back office and mid-off processes being worked upon with AI and it's really elevated the customer experience. It's reduced the cost in a very meaningful way and the client is working with us in other areas to help them embed AI into the workflow.
Those results are phenomenal. I'm glad you mentioned customer experience because listening to what you're saying I mean the organizations probably had a lot of productivity and you know other benefits in mind but I think one of the real wins here is how fast you're able to serve that customer right and get answers faster be more responsive can you talk a bit about that?
