The Missing Link in Healthcare AI: Domain Expertise + Workflow

Welcome everyone. Welcome to today’s edition of HFS Unfiltered. And today we are going to talk  

0:28

28 seconds

about everything in what’s happening in the U.S. healthcare. You know, it’s a fascinating space.  

0:21

Chapter 1: U.S. Healthcare’s Trillion-Dollar Challenge

21 seconds

0:35

35 seconds

On one hand, the U.S. healthcare administration is almost a trillion-dollar drag on the  

0:43

43 seconds

U.S. economy. The payer MLR pressure is brutal right  now. The provider margins are thin. And you know,  

0:52

52 seconds

on the other hand, the age-old BPO and  the outsourcing world of more FTEs,  

0:59

59 seconds

lower cost is plateauing, and as we  coined services as software in 2024,  

1:07

1 minute, 7 seconds

that whole market of AI-enabled solutions  is starting to kick in. So to discuss all  

1:13

1 minute, 13 seconds

the headwinds and tailwinds, I thought who  better than to talk to the CEO of Sagility,  

1:22

1 minute, 22 seconds

one of the biggest pure play healthcare business  services companies, Ramesh Gopalan, on what his  

1:30

1 minute, 30 seconds

thesis is on the healthcare market, what Sagility  is doing, and really deliberate on the services  

1:38

1 minute, 38 seconds

as software economy and their bet on what they’re calling Synchrony. So Ramesh, welcome to the show. 

1:47

1 minute, 47 seconds

Thank you Saurabh, good to see you. So Ramesh, let me start by asking you, what do you  

Chapter 2: The Biggest Challenges Facing Payers and Providers

1:56

1 minute, 56 seconds

think is the single biggest operational failure  point that you’re hearing from the CXOs of both  

2:04

2 minutes, 4 seconds

payer and provider clients that you’re talking to? Sure, to just give you a context, and you alluded  

2:11

2 minutes, 11 seconds

to some of this in your opening remarks. U.S. healthcare is going through a very tough period,  

2:19

2 minutes, 19 seconds

if I may say that, right? So there is a huge  profitability pressure both on the payers and the  

2:24

2 minutes, 24 seconds

providers. MLR, like you spoke about, is pretty  high for most of the payers, especially those  

2:31

2 minutes, 31 seconds

who operate in the government sponsored plan  space, and it’s something that’s a big worry  

2:41

2 minutes, 41 seconds

for all the healthcare participants, both payers  and providers, right? So that’s something that  

2:47

2 minutes, 47 seconds

CXOs are trying their best to wrap their  arms around, and the bigger participants,  

2:55

2 minutes, 55 seconds

as you know, have done traditional outsourcing  for a number of years. I mean, we’ve had clients  

3:01

3 minutes, 1 second

who’ve been with us since 2000. So the larger  players have done this for 20-25 years. They’re  

3:07

3 minutes, 7 seconds

pretty mature outsourcers. They’ve taken advantage  of all that labor arbitrage could give them,  

3:14

3 minutes, 14 seconds

but now they need more. So it’s that which is  a struggle, and the whole promise of AI and  

3:23

3 minutes, 23 seconds

what it can give is something that everybody  is keen to explore, but at the same time,  

3:30

3 minutes, 30 seconds

they’ve just scratched the surface, and some  of them who’ve run pilots, they feel that  

Chapter 3: Why Most AI Pilots Fail to Scale

3:36

3 minutes, 36 seconds

they’re not getting the ROI that people promised  them. And so we are at that stage where there is  

3:42

3 minutes, 42 seconds

a promise of AI, people are desperately looking  to take cost out of the system, and still trying  

3:49

3 minutes, 49 seconds

to figure out what’s the best way to do that. Yeah, Ramesh, you’re so right about the whole  

3:55

3 minutes, 55 seconds

conversation on pilots. I often joke that AI  is going through a death by a thousand POCs  

4:02

4 minutes, 2 seconds

right now, where everybody is running a POC in one nook and corner of their organization,  

4:08

4 minutes, 8 seconds

but very few of these POCs are scaling up.  I know you’re making a big bet on AI-led  

4:16

4 minutes, 16 seconds

solutions or what we call services as software with Sagility Synchrony. Tell us a little bit  

4:23

4 minutes, 23 seconds

about what that does in a typical healthcare workflow and how is it different, and how does  

Chapter 4: Sagility Synchrony and AI-powered Workflows

4:30

4 minutes, 30 seconds

it address some of these problems of scaling AI? Good question Saurabh. So if you look at the  

4:38

4 minutes, 38 seconds

reason why some of these pilots have not proven  successful, it’s because healthcare is complex,  

4:44

4 minutes, 44 seconds

right? So it’s very fragmented, and there are  multiple workflows, all of which use multiple  

4:51

4 minutes, 51 seconds

systems, pretty disparate decision making. And so  if you attempt to use AI to just automate steps  

5:01

5 minutes, 1 second

in a process, that’s a journey that’s not going  to yield the ROI that you require. So when we  

5:08

5 minutes, 8 seconds

looked at these fragmented workflows, and we’ve  been in this business for 25 years, we have deep  

5:14

5 minutes, 14 seconds

domain expertise on the U.S. healthcare side. And  so we are looking at the critical workflows,  

5:22

5 minutes, 22 seconds

for example the whole front-end enrollment,  sales and enrollment workflow, or the prior  

5:28

5 minutes, 28 seconds

authorization workflow, and we are looking at some  of these broken fragmented workflows, and we’re  

5:34

5 minutes, 34 seconds

trying to see how do you do an orchestration using  AI with domain rules and workflow automation,  

5:42

5 minutes, 42 seconds

all bundled in with human expertise also on the  same operational layer. And that’s what we are  

5:47

5 minutes, 47 seconds

calling Synchrony. For example, if you take the  sales and enrollment along with partner platforms,  

5:58

5 minutes, 58 seconds

we are looking at the whole value chain starting  from bid generation, document submission to CMS,  

6:06

6 minutes, 6 seconds

followed by broker and sales management, and then  compliant enrollment and billing and so on. How do  

6:13

6 minutes, 13 seconds

I look at that overall workflow and how do I bring  in, like I said, domain rules and AI intelligence  

6:19

6 minutes, 19 seconds

and humans on the same operational layer so  that you’re no longer looking at activities  

Chapter 5: Reducing Costs While Driving Revenue

6:26

6 minutes, 26 seconds

and automating distinct steps in the process,  but you’re looking at an end-to-end workflow  

6:30

6 minutes, 30 seconds

and you’re trying to optimize the outcomes. In  this case we think you can easily get a 30-40%  

6:37

6 minutes, 37 seconds

reduction in both technology and operational cost,  but more than that, the more important metric is  

6:44

6 minutes, 44 seconds

how do you improve the revenue uplift, right?  So if you can enroll people more compliantly,  

6:50

6 minutes, 50 seconds

you reduce the capitation timing loss, you have  a better RAF score, lower premium billing errors,  

7:01

7 minutes, 1 second

all of this adds up significantly, right? So  that’s the way we’re looking at Synchrony.  

7:07

7 minutes, 7 seconds

And what it also does in a relationship between a  payer and a service provider like us, is payers no  

7:14

7 minutes, 14 seconds

longer have to pay for activities, right? So the  traditional outsourcing model, you give a piece of  

7:19

7 minutes, 19 seconds

work, it’s all kind of orchestrated in the systems  of the payer, and they pay you by effort. Either  

7:29

7 minutes, 29 seconds

it’s an FTE rate or a transaction rate, and that  was the traditional model, whereas here you pay  

7:34

7 minutes, 34 seconds

for outcomes. So the outcomes in most cases you  can price on a per member per month. So it’s a  

7:41

7 minutes, 41 seconds

pretty variable cost for the period. And so they  know exactly what it is that they are spending.  

7:47

7 minutes, 47 seconds

And as a service provider, we tried to optimize  by reducing reworks, by reducing the number of  

Chapter 6: Moving From Activities to Outcomes

7:55

7 minutes, 55 seconds

touches, and basically using the power of AI  to drive some of these end-to-end workflows. 

8:02

8 minutes, 2 seconds

So Ramesh, that’s very exciting, and there are  two things that caught my attention. One was,  

8:09

8 minutes, 9 seconds

it’s not just about cost. Yes, cost has  to be a big driver of it, but you’re also   looking at the revenue uplift angle, which is  fantastic, because you know costs at the end  

8:20

8 minutes, 20 seconds

of the day will be a diminishing return. You  know how much lower can you ultimately get.   And the other thing that caught my attention  was the whole outcome-based pricing model,  

8:30

8 minutes, 30 seconds

which practically is another one of those  enigmas where everybody promises outcome-based  

8:36

8 minutes, 36 seconds

and nobody’s able to deliver on that. But I think  what you’re talking about, Sagility Synchrony,  

8:45

8 minutes, 45 seconds

seems like a completely different operating  model as well. So it’s not just some AI solution,  

8:50

8 minutes, 50 seconds

but it feels like you’ll need to rethink your  sales motions, your delivery organization,  

8:56

8 minutes, 56 seconds

how you price deals. You know the whole narrative  around Sagility needs to change. It’s not just  

9:05

9 minutes, 5 seconds

some new tool that’s out there. It’s a big  shift. Is that a fair observation, Ramesh? 

9:14

9 minutes, 14 seconds

Absolutely right. So it requires change all  through the organization. You’re absolutely  

9:20

9 minutes, 20 seconds

right. So look, we’ve done the healthcare  services business for 25 years. We’ve gained  

9:26

9 minutes, 26 seconds

the domain expertise. But if you historically  look at how did that expertise manifest itself  

9:32

9 minutes, 32 seconds

in the value that we deliver for clients, it’s  process improvement. So again, piece by piece,  

9:38

9 minutes, 38 seconds

you take a process, you try and optimize it,  you do business transformation by optimizing the  

9:45

9 minutes, 45 seconds

process, you introduce automation in steps of the  journey, and then you also have point solutions,   right? These were the traditional ways by which  you brought your expertise and added value to  

9:54

9 minutes, 54 seconds

the client. But today we are not looking at  it as a single or a few steps in a process.  

10:00

10 minutes

When you look at end-to-end workflow, it’s no  longer task execution, but it’s more a workflow  

10:05

10 minutes, 5 seconds

execution. So that requires a different way that  you bring the delivery folks to the table. That’s  

10:11

10 minutes, 11 seconds

a different way you apply the domain intelligence  to the outcomes. And also the sales motion, you’re  

10:18

10 minutes, 18 seconds

absolutely right. How you go to the market, how  you position the whole Synchrony value offering,  

10:24

10 minutes, 24 seconds

and how you price those deals, all of that needs  to change. And so, all of those changes while it’s  

10:30

10 minutes, 30 seconds

happening, we are investing a lot in building the  Synchrony suite of solutions. It’s reusable IP,  

10:37

10 minutes, 37 seconds

but it’s not just done by us alone. So we are  dependent on a partner ecosystem. We’re working  

10:45

10 minutes, 45 seconds

with some platform players who are also trying  to introduce agentic workflows in their software  

10:52

10 minutes, 52 seconds

solutions. So it’s a whole change in the way we  go to market, in the way we craft these solutions,  

10:58

10 minutes, 58 seconds

and also the delivery organization that’s going  to stand behind it and deliver the outcomes. 

Chapter 7: Advice for Healthcare Leaders Adopting AI

11:04

11 minutes, 4 seconds

Yeah. So Ramesh, let’s now flip this. So you  talked very eloquently about how your operating  

11:10

11 minutes, 10 seconds

model needs to change and is changing. I think  the buyers, or your client’s operating model,  

11:16

11 minutes, 16 seconds

will also have to change, right? Otherwise  the issue becomes it will be restricted to  

11:24

11 minutes, 24 seconds

yet another POC and a pilot which doesn’t scale,  unless they change their operating model as well.  

11:30

11 minutes, 30 seconds

So imagine if let’s say there is a healthcare  payer COO listening to this conversation  

11:38

11 minutes, 38 seconds

who has a $200-300 million IT outsourcing BPO  portfolio of vendors, hopefully Sagility included  

11:49

11 minutes, 49 seconds

in it. What’s your advice to them? What’s the  one thing that they should start doing, and  

11:56

11 minutes, 56 seconds

what’s the one thing that they should stop doing? So that’s a good question. While I spoke about how  

12:02

12 minutes, 2 seconds

we are changing, I also want to say that look,  you can’t be the same to every client. So there  

12:08

12 minutes, 8 seconds

are large clients who still want to do some of  this internally. So our positioning to them is,  

12:14

12 minutes, 14 seconds

hey, let’s be your AI enablement partner. We  understand the process. Let’s redesign your  

12:21

12 minutes, 21 seconds

workflows. Let’s help you build the business case,  and let’s help you implement these solutions.  

12:28

12 minutes, 28 seconds

Maybe they won’t transition overnight to a fully  outcome model, but at least we can help you  

12:34

12 minutes, 34 seconds

through the journey of redesign of workflows. But then when you look at the mid-market and so on,  

12:39

12 minutes, 39 seconds

Synchrony as a full suite of solutions may be more applicable to some of those clients. So we’re also  

12:45

12 minutes, 45 seconds

looking at it differently for different sets of clients. And to the other question you asked,  

12:51

12 minutes, 51 seconds

I mean typically the same conversation that we had around pilots. So, our recommendation to client is,  

13:00

13 minutes

don’t just go about trying to apply AI to every problem. Take a look at workflows. All  

13:08

13 minutes, 8 seconds

players have these workflows, be it around prior authorization, be it around claims, be it around enrollment. Look at the workflows where you have the most leakage, right?  

13:18

13 minutes, 18 seconds

Number of touches, avoidable cost, and identify those workflows, and try and implement a Synchrony  

13:26

13 minutes, 26 seconds

kind of a solution. Look at it end to end, use AI-based orchestration, try and identify what  

13:32

13 minutes, 32 seconds

are the metrics that you’re going to impact, and then bring in the domain expertise and the humans  

13:39

13 minutes, 39 seconds

in the loop, and try that out. Don’t just try to automate steps in a process. Yes, you will  

13:45

13 minutes, 45 seconds

get some benefit, but it’s nowhere near the same ROI that you’re expecting. And the other thing is,  

13:51

13 minutes, 51 seconds

I mean don’t go about replacing one large platform with another, because that’s not going to solve  

13:57

13 minutes, 57 seconds

your problem. So that’s something that we talk to our clients. But then again, like I said,  

14:04

14 minutes, 4 seconds

I mean different clients have different approaches, and so our hope is we can be a  

14:09

14 minutes, 9 seconds

partner to them either in the design and help them implement, or if they want us to take control of  

14:17

14 minutes, 17 seconds

a workflow end to end, we can be the ones taking ownership and being accountable for the outcomes. 

Chapter 8: Why AI Can’t Fix Broken Processes

14:24

14 minutes, 24 seconds

Yeah, you’re so right. I think with every technology intervention, and I’ve seen this  

14:31

14 minutes, 31 seconds

throughout the 30 years that I’ve been in this industry, you know, a new technology  

14:36

14 minutes, 36 seconds

comes in and we use the technology to do the same thing maybe slightly faster, slightly cheaper,  

14:44

14 minutes, 44 seconds

and then we say where is the transformation? And unless we resolve some of the process debts that  

14:52

14 minutes, 52 seconds

you articulated around the workflows, whether it’s AI or some other magic bullet, it’s not going to  

15:00

15 minutes

solve all your problems. We have to look at those.  So, I think we’ve talked a lot about scaling AI and  

Chapter 9: The Future of Talent in Healthcare

15:11

15 minutes, 11 seconds

outcome-based pricing, operating models, etc. One of the other big questions, Ramesh,  

15:16

15 minutes, 16 seconds

that’s swirling around us is the impact on talent. You know at the end of the day our  

15:22

15 minutes, 22 seconds

industry is a people industry. Services industry has always been a people industry, and I’ve seen  

15:29

15 minutes, 29 seconds

all kinds of geometric shapes being thrown around now when it comes to the talent model.   You know it’ll be a K-shaped model, it’ll be a T-shaped model, it’ll be a diamond shaped model,  

15:41

15 minutes, 41 seconds

and all kinds of things. Where do you think we are headed as an industry from a talent perspective?  

15:51

15 minutes, 51 seconds

If there’s a young guy or gal listening to us, what are their prospects in the future? Let’s say  

16:00

16 minutes

if Sagility Synchrony, I hope and wish it’s very successful and you meet your ambitions, if agentic  

16:07

16 minutes, 7 seconds

systems are going to replace the human workflow, where do we go from a talent perspective? 

16:16

16 minutes, 16 seconds

Like I said, healthcare is complex. So, while we’re talking about agentic systems, we don’t think  

16:22

16 minutes, 22 seconds

any of these workflows will be done purely by agentic systems. There’s always a place for human   in the loop. There is judgment, there is clinical nuance, there is regulatory constraints. There are  

16:33

16 minutes, 33 seconds

a number of reasons why by design you need to have human in the loop at different points in time. And  

16:39

16 minutes, 39 seconds

when we design Synchrony, that’s a core part of the workflow design, trying to figure out what’s  

16:46

16 minutes, 46 seconds

the right place for humans to intervene. And also, I mean these agentic workflows are not going to be  

16:53

16 minutes, 53 seconds

cast in stone. So you need to constantly monitor the outcomes, and you need to constantly keep  

16:59

16 minutes, 59 seconds

improving, which will again require human involvement. So, the way we look at it is,   look, domain expertise is going to be key, and so that’s what we tell our people. It’s no longer  

17:09

17 minutes, 9 seconds

about following a set of instructions and doing things over and over again. It’s  

17:14

17 minutes, 14 seconds

how do you bring domain intelligence into play, and how do you elevate your expertise  

17:20

17 minutes, 20 seconds

with the help of AI to deliver exponential outcomes? So, I think the roles will change,  

17:26

17 minutes, 26 seconds

more expertise will have to find its place, but there’s always going to be humans in the workflow.  

17:36

17 minutes, 36 seconds

And so we think there is still a place for domain experts and people who understand the industry.

17:43

17 minutes, 43 seconds

Well, that’s well said. I think it’s about elevating all our jobs and essentially solving the industry problems. It’s not about replacing people  

17:55

17 minutes, 55 seconds

with AI. So, Ramesh, this has been a fantastic conversation, but before I let you go,  

18:01

18 minutes, 1 second

if I had the power to grant you one wish, and let’s keep it to the healthcare industry,  

18:08

18 minutes, 8 seconds

and let’s just talk about one wish. What would be the one wish that you hope comes true this year?

Chapter 10: The Vision for Agentic AI at Scale

18:17

18 minutes, 17 seconds

Look, there’s so much potential, right? And there’s so much promise of what AI can do. I just  

18:24

18 minutes, 24 seconds

wish that the industry as a whole, both clients and us, are successful in implementing it, and  

18:33

18 minutes, 33 seconds

getting the power that these agentic workflows can  get. So I definitely think it’ll happen. I think  

18:42

18 minutes, 42 seconds

it’s sometimes a question of in what time frame.  So people wish it’ll happen in the next 3 to 6  

18:48

18 minutes, 48 seconds

months. I think given the complexity of the industry, it’s probably going to take longer. But  

18:53

18 minutes, 53 seconds

I think in the next 3 to 5 years, I’m very hopeful that at least 50% plus of the workflows in any  

19:03

19 minutes, 3 seconds

large healthcare organization will be orchestrated through agentic systems, and that’s the hope.

Chapter 11: Closing Thoughts

19:12

19 minutes, 12 seconds

Fantastic, Ramesh. With that wish and the hope, thank you again for spending some time   with us. I really enjoyed this conversation, learned a lot about the healthcare industry,  

19:23

19 minutes, 23 seconds

what Sagility is doing, and where we think we can move with   realizing the promises of AI. So, thanks a lot Ramesh for spending some time with us.

19:34

19 minutes, 34 seconds

Thanks Saurabh. Thanks for having me on your show.