Revenue Cycle's AI Moment: A Conversation with Jacob Shurbet, Principal, PwC US — June 5, 2026

Healthcare organizations are under mounting financial pressure, and the revenue cycle sits at the center of it. Billing complexity, payer friction, labor shortages, and fragmented data systems are converging to suppress realized revenue at scale. Jacob Shurbet, a principal at PwC, has spent years working with health systems on the operational and strategic dimensions of that challenge. 

In a recent BizTechReports executive interview, Shurbet laid out how AI is changing the terms of the problem, why most organizations are still struggling to capture the value, and what it actually takes to move from experimentation to execution.

Here is what he had to say:

BTR: The revenue cycle management in healthcare has been around for decades. Why is it suddenly a C-suite conversation?

SHURBET: Maybe I'll start by actually rewinding 30 plus years. The revenue cycle was in existence, but still very manual, paper-based, very much focused on the transactions required to collect cash. Lots of phone calls, paper workflows moving between different people that were part of the process. 

Then in the early 2000s, bolt-on workflow tools started to become more ubiquitous. Providers started adopting elements they were layering onto their core EMR and patient accounting platforms to help digitize a lot of that paper flow. About ten years ago, that's when automation started to come into play. Screen scraping, robotic process automation, more basic automation, but it still had a place. Digital transactions started to occur through remits, through HIPAA interoperability requirements, and that really drove a lot of adoption to improve how information was flowing between all the parties involved. 

And then you land where we are today, which is where AI is driving every conversation we're part of. It is moving extremely quick. But the reason it's become a C-suite conversation goes back to something fundamental. Your revenue cycle is a continuum of operations, and you have to get everything right for it to actually work. With the volume of transactions that flow through it on a daily basis, that's really hard to do. Oftentimes providers miss core components that are required to have the right focus on, so that you're not ending up at a place where your back office is cleaning up mistakes that occurred earlier in the process. When the financial stakes are this high and the margin environment is this difficult, that belongs on the leadership agenda. 

BTR: Everyone talks about AI as a cost-cutting tool. Is that the whole story?

SHURBET: Most people quickly go to reduce cost. Yes, I think that's a true statement, but there's also an upside and an additional yield component that doesn't get nearly enough attention. 

Think about how complex the revenue cycle is from the point when a patient walks in the door. You have to verify information, capture codes and documentation on the service, submit the claim, have the payer review it and make sure it matches up with your contracts and agreements and all of the line items that dictate what you should get paid for that service. Then you have to get all that information back, post it to the right account, and get that account to zero. You can imagine all of the mistakes that can get made along the way. And then the deeper question is, am I actually getting paid what I should based on the complexity of my contracts and the services I provide? 

Healthcare billing isn't like walking into a retail store and buying a t-shirt. There's a lot more to it in terms of how prices are set, how they're loaded in the system, how they're maintained, how they flow through when you actually bill the payer, and how the payer can pay what they think is accurate but oftentimes isn't. 

So when I think about where AI drives real upside, think about just contracts and payment integrity. Using AI in a way where you can start to scan large amounts of data and transactional information and find gaps. I have a payer that historically I've always thought has been paying fairly accurately. Maybe there's a slight differential, small percentage points here or there. Over time though, that adds up if you're not keeping track of it with the number of patients these healthcare providers see. Layering on AI in a sophisticated way, building a reimbursement defense application that monitors transactions at the most granular level, that is where I see AI driving a lot of benefit from additional revenue capture. Recovering that revenue is a precision problem, not a volume problem. 

BTR: You cited data from the Healthcare Financial Management Association that stopped me cold. Can you share that?

SHURBET: This was actually reported last year from HFMA, the Healthcare Financial Management Association, which is a big player in the healthcare finance space. They said 88% of health systems are using AI in some form, but only 18% have a mature governance structure and fully formed AI strategy. That paints a pretty clear picture of where the fragmentation actually lives. 

Everybody wants to use it, everybody's trying to use it, but there's still a lot of fragmentation of the how and the what. And I think organizations are missing the point that it starts with governance and it starts with a strategy and a clear vision of what it's gonna take to actually achieve that ROI. That's where organizations are still struggling today. You end up with point solutions that generate localized results but never move enterprise-level metrics. Days in accounts receivable don't move. Net collection rates don't move. And leadership starts to wonder whether any of it is actually working. 

The gap between 88% and 18% is where most of the frustration lives. And until organizations close that gap, the technology is going to keep underdelivering relative to its actual potential. 

BTR: You drew a sharp distinction between automation and orchestration. What does that mean in practice?

SHURBET: Historically, when you think about automation, it was about bits and pieces of a process. You'd look at a process map, identify the steps that could be automated, build decision trees around them, and automate those components. The unit of work was the task. 

We're saying something different. What is the outcome you're trying to get to? And then how do we actually build an agent to achieve that outcome with a human in the loop where necessary? That's the paradigm shift. And it matters because revenue cycle workflows contain constant variation. Every claim reflects different clinical conditions, different payer rules, different contract terms. A fixed workflow can't handle that. AI agents can assess context, determine next actions, and coordinate across systems. The process adapts to the claim rather than forcing the claim through a predetermined path. 

Think about resolving a denied claim without human intervention. Or ensuring that every service rendered is coded, billed, and reimbursed correctly the first time. That's outcome accountability, not task execution. And it changes how work is distributed across both technology and people in a pretty fundamental way. Organizations still thinking about automation as a list of tasks to eliminate are already behind. 

BTR: How does PwC’s recently announced collaboration with AWS fit into that thinking?

SHURBET: It started a little over a year ago when we were building out our five-year roadmap as a business, thinking about what we knew we needed to accomplish given where we saw technology and AI going. We did a scan, talked to several different organizations. PwC at large has relationships with Microsoft, Google, AWS, and others. We felt like AWS's presence in the healthcare space, some of the proprietary tools they built and how they were managing different aspects of the provider data ecosystem, along with Amazon Connect, which they had deployed at a lot of providers, fit the profile as a good foundation we could build upon. Not only with their data lake and what they had in place for managing data, because data is the ground zero for AI, but also because Amazon Connect had real potential to build upon relative to how it was already being deployed at providers. 

Our premise was to reimagine the revenue cycle. We don't want to stop and think about what could be done based on what has always been done. Throw that out the door and start over. Reimagine it altogether of what it could look like five years from now. 

PwC brings domain expertise in revenue cycle operations, process redesign, and the governance frameworks required to operationalize AI at scale. AWS brings the cloud infrastructure, data architecture, and AI services needed to aggregate, process, and act on large volumes of healthcare data securely. Together the integration enables a unified operating model that connects fragmented systems and allows AI to drive continuous, end-to-end improvement across the revenue cycle. 

The way we're thinking about it is, how do you take all of that data, build an AI orchestration layer, which is essentially the housing for your AI agents, where all the data flows through to feed your agents on what needs to happen, and then flows back into all the systems that work in conjunction to make your business run. Most organizations aren't thinking about it that way yet. They're still thinking about individual vendor solutions for individual problems, and none of that talks. None of that does the job it's supposed to do to get to the five-year vision. 

BTR: What happens to the workforce as agents take on more of the transactional load?

SHURBET: We think about it in three phases, and we've been very intentional about that framing because it's not something that just magically happens tomorrow. 

Phase one is augmentation. We're really looking at using AI to empower our staff to do better, to do things faster, to help them move from point A to point B along the continuum of how they're doing their work, with the support of AI. A good example is what we call next best step action. That's surfacing relevant information in real time as a staff member is pulling up an encounter and a claim, helping them understand what steps they should take to properly resolve it. When you're working accounts receivable, working open claims and denials, the largest portion of time spent is the research. What is the payer asking for? What documents do I need to gather? What actions do I need to take? That's usually the most time-intensive part of actually getting to a point where you can process a claim and move it toward payment. Phase one is about compressing that research phase significantly without replacing the human doing the work. 

Phase two is where the work itself starts to change. Agents are handling more of the transactional load, and staff shift toward managing complexity, handling exceptions, and overseeing agent-driven workflows. Their work's going to change. We're going to have to rethink about the role they play. Agents can read a contract and tell you what's in there, applying the concepts and surfacing the relevant information. But actually having an interaction or a negotiation with a payer, agents aren't going to be able to do that just yet. Your staff are going to have to take that information and apply it in the right ways. The critical thinking, the judgment, the negotiation, that stays human. What changes is everything underneath it. 

Phase three is where humans are now building, configuring, and refining the agents themselves. It's not about actually doing the work anymore. It's about how do I keep the agents up to speed and in line with what I need them to do. That's a significant shift in how we think about training and workforce development, and it's something we're planning for now even though the timeline isn't fixed. Five years is our best guess. Quite frankly it could be sooner with some of the advancements we're seeing on a regular basis. 

BTR: Your firm's own 2026 global CEO survey found that only 12% of CEOs say AI has delivered both cost and revenue benefits. What's the diagnosis?

SHURBET: It's not only an AI problem. It's a cultural problem. It's a data problem. It's an adoption problem. And I think they are missing the point that it starts with execution and it's not just the technology. We know the technology is there and we know its potential. But if you're not executing in the right way, and if you're thinking about it more narrowly than you should, you're never going to have the success organizations are hoping to achieve. 

We've seen this dynamic before. When we came in to help clients with EMR adoption, it wasn't that the technology couldn't do what it was supposed to do. Organizations didn't think about the adoption and the organizational change that has to happen to really unlock the value. That's no different in the phase we're in with AI. It's just happening a lot faster and at higher stakes. 

56% of CEOs say they've seen no significant financial benefit to date despite the experimentation. That's a staggering number given how much has been invested and how much organizational energy has gone into it. A lot of providers are still thinking about it very myopically, use case by use case, rather than taking a step back and asking where could we be three years from now and how do I get there phase by phase. The POC hell problem that gets reported on is real. You end up with a collection of isolated pilots that never scale, a lot of wasted effort, and a growing sense that maybe this isn't going to deliver what everyone said it would. That's a strategy and execution failure. 

BTR: Walk me through what the framework for getting this right actually looks like.

SHURBET: It starts with strategy and vision. Not just I want to lower cost and improve revenue. Everyone knows that. There has to be more specificity around what you're actually trying to achieve and how you're going to get there. The exciting part is that most organizations we speak to, when we talk to them about this, revenue cycle is the area they want to go after first because it's so transactional and they feel like there's a lot of value there. But even within revenue cycle, what are you really struggling with the most today? It doesn't have to be I want to automate my entire revenue cycle. It can be certain parts you need to tackle first. That doesn't change the approach though of how you need to go about it. 

Then it's the data. Getting that orchestration layer built, getting everything in place, getting all the tools set, codifying it, making sure you understand how the inflow and outflow is going to work. Your EMR, your patient accounting system, your clearinghouse, your contract management tools. All of them are full of data that isn't talking to anything else. Some organizations have started to bring that data to the cloud and think about how they're leveraging it. Most are still falling short of where it could be. That has to change before AI can operate as a system-wide capability rather than a point solution. 

Then there's responsible AI. We've got to test it, make sure it's compliant, make sure it doesn't hallucinate, make sure it fits within the guardrails. Especially in healthcare, depending on how you're applying it, that's not optional. The risks are real and you have to be very thoughtful about how you're managing them. 

And then you've got to measure it. Going back to the old Six Sigma model, measure what matters. The traditional revenue cycle metrics don't change. Days in accounts receivable, net collection rates, cost to collect. If what you're doing is working, those numbers move. You're not inventing brand new metrics that suddenly point you in the right direction. You're watching the metrics you've always measured improve, and by what clip you would expect them to improve. That's how you know. 

BTR: Give me a concrete example of what execution looks like when all of that comes together.

SHURBET: In about 10 weeks, we built and deployed a conversational AI agent that's able to make outbound calls to payers and handle those calls mostly autonomously. I say mostly because we're still refining it, but for the most part it can call, it can interact just like you and I are interacting here. It can ask the questions, it'll prompt with the information necessary to get through the different tollgates. The agent will ask for the patient's date of birth, the patient name, the claim number, it can do all of that. And then it'll start to actually ask questions based upon the information we fed it around why a claim has been denied or why it hasn't paid. It'll generate very thoughtful questions, ask clarifying questions back to the human on the other line, capture that information, and plug it back into the system. 

That framework I described, we went through all of it with that one use case. The data integrity, the responsible AI component, the compliance requirements, the training. All of it had to be right before we could deploy. And we did it in 10 weeks. Going from concept to deployment in that timeframe, that just shows you the potential. Early successes like that are starting to shift expectations around what can actually be automated and how quickly value can be realized. 

BTR: Last question. What's the message for leaders who are still waiting to see how this plays out?

SHURBET: AI is not just an easy button. You've got to have a good strategy, a good plan, a good data foundation, and then actually go after that. And I think if organizations are well positioned and understand what's coming, they're going to be able to capitalize on it. The ones that aren't are going to find themselves behind. 

I'm in the room having these conversations with providers every day. And I can tell you that the ones still thinking about it use case by use case, waiting for the technology to mature a little more before they commit to a real strategy, those are the organizations that are going to look up in three years and wonder how the gap got so wide. The advancements we're seeing on a regular basis are accelerating. And the distance between organizations that are executing well and organizations that are still experimenting is going to become very visible very quickly.