November 2026—The laboratory can be seen as the diagnostic ecosystem for the entire enterprise. As such, the laboratory information system, or LIS, is the CNS, or central nervous system, said Liron Pantanowitz, MD, PhD, MHA, of UPMC/University of Pittsburgh. “It’s the digital backbone to everything.”
Dr. Pantanowitz spoke online with CAP TODAY publisher Bob McGonnagle and others about the LIS: putting all laboratories in a health care system on a single LIS platform, going paperless, deploying AI, and making business-related decisions that rely on the data the LIS generates.
“My wish for the LIS is to provide even better data analytics tools,” Dr. Pantanowitz said.
S. Joseph Sirintrapun, MD, of Mass General Brigham, and Omar Baba, MD, of Henry Ford Hospital, joined Dr. Pantanowitz and industry representatives on Sept. 22 for the LIS close-up led by McGonnagle. Their conversation follows.
Click here for CAP TODAY’s interactive Laboratory Information Systems product guide.
The laboratory information system gets less respect than it deserves. We have terrific AI technologies, including an important cytology system, as well as EHRs, applications, and middleware. Joe Sirintrapun, what is your current experience with the LISs at Mass General and how are you interfacing with them?
S. Joseph Sirintrapun, MD, associate professor of pathology, Harvard Medical School, and clinical director of digital pathology, Mass General Brigham: Mass General Brigham, which includes Mass General Hospital and Brigham Women’s, were on different LISs. We wanted to transition the entire vast network, which spans 10 hospitals, including two large academic medical centers and several community/health center labs. We had been on an Epic EHR system, and we launched Epic Beaker last November. It was a large effort.
Standardizing is not an uncommon story today. Most of us are floating on an ocean of Epic IT, both in the EHR and LIS. Nick Trentadue, tell us about your efforts in Boston.
Nick Trentadue, vice president, laboratory and diagnostics, Epic: Epic’s mantra from the get-go has been to focus on three key segments: pediatrics, large integrated delivery networks, and academic medical centers. We had a later start on the lab side. Beaker came out in 2006, and for years I told groups like Mass General we were not ready for them. Over the past four to six years, we’ve been coming back and talking to groups like UPMC, Mass General, and others to say it’s time to take another look at Beaker. Our R&D team has worked with customers to make the product better, and we’ve redeveloped the entire pathologist workflow. That effort by R&D and engineering is what you’ve started to see with the academic medical centers that have completed their transition, along with several others in the midst of it now.
We have more work to do to enhance the experience. We’re integrating AI and models and predictions into Beaker, and we look forward to developing more functionality.
Suren Avunjian, LigoLab has a bit of a different story as a laboratory information system. Tell us what your focus is.
Suren Avunjian, cofounder and CEO, LigoLab Information Systems: Our focus has been helping independent pathology groups and clinical labs that are scaling. They are adding clients, specimens, and often more complex services while still competing on turnaround, quality, and responsiveness. Our platform brings the laboratory and its financial workflow together, from order and accessioning through reporting, billing, and follow-up, in one system. The lab team spends less time reentering data or tracking down the next hand-off. We can automate routine work and surface exceptions for the people who need to resolve them. That lets a lab handle more volume with the team it has. It also helps the lab offer services and reporting that fit different clients and distinguish it in the marketplace.
Omar Baba, Henry Ford’s strategy has been to grow, adding capacity and laboratories. Does your LIS need to ride along with you to support that initiative?
Omar Z. Baba, MD, clinical pathologist, pathology informaticist, and currently clinical pathology resident, Henry Ford Hospital: Yes. It has long been a goal of our leadership to bring any laboratory that joins the Henry Ford family onto the same LIS. That speaks to our centralized model and the standardization we push for. It’s not easy to bring everyone on the same system; integration takes a lot of work.
Liron Pantanowitz, UPMC is a very large system. Tell us your thoughts about being on a standardized platform.
Liron Pantanowitz, MD, PhD, MHA, chair and Maud Menten professor of pathology, UPMC/University of Pittsburgh: Our CEO just acquired Ohio-based Trinity Health system, so we are now up to 46 hospitals in our system and initially they all had labs. We had seven electronic health record systems, so everyone needed to get on the same EHR, and as we buy more, there will be more. Epic was the enterprise choice; the LIS module is Beaker. I was not happy we’d have to endure the transition of all the different LISs, but I was happy we could now have an LIS of almost everything. The only one we’re not incorporating is the blood bank.
From an IT perspective, the CIO tells me it’s more than the standardization; it’s a simplification of infrastructure. They love that they will be supporting just one application. It was a catalyst to look at our data centers because many legacy systems don’t have to be supported now.
Given the size of our operation, we split the transition to Beaker into waves. We chose wave one for our peripheral community hospitals and wave two for the more complicated academic hospitals around Pittsburgh. While the stepwise approach was beneficial, to learn from the mistakes or best practices of wave one, it was tough to have a hybrid LIS adoption within the same system. One or two hospitals within the same system were caught between waves one and two and were not on Epic. They would send their specimens to the hospital in their region to get grossed and have slides made, and then they were sent back because they were not fully digital. They were working out of two LISs. That was tricky.
We also realized it’s impossible to train everyone to use Epic Beaker because we have so many different tasks and niche jobs. For example, there’s no training in the LIS that will apply to a biochemical geneticist.
We’re going live in a week or two, and then we’ll have thousands of users. Our major concern is to not miss capturing any revenue.
From a change management perspective, the hardest thing has been going paperless at the same time.
Chris Malek, does this sound familiar to you as you sit in on client meetings and hear where customers and potential customers are headed?
Chris Malek, senior manager, product management, Clinisys: Yes, especially about the challenge to break free from paper and go to a more electronic, automated system. But we are seeing people embracing it and trusting it more, especially with AI coming into play.
Regarding standardization, having a central system that has all your disciplines together is great for bridging gaps between clinical and anatomic pathology. With AI, it will be critical to have all your information collated, organized, and standardized in one system, with integration bridging gaps if needed. That will be critical for the AI features and functionality coming out.
Nick Trentadue (Epic): It pains me a little when a group says, “We’re going digital but we want the cover sheet.” Not only can we give you the cover sheet electronically every time you open the chart, but you can tune your AI to tweak it further in real time. Do you really want this piece of paper? I’d say 50 percent still do.
Joe Sirintrapun, does this all sound familiar to you at Mass General Brigham?
Dr. Sirintrapun (Mass General Brigham): Yes. Less than a year into our go-live, we’re still trying to optimize. It was good to standardize across the network. We gained a lot, but many things were hard to foresee. There were also transitions involving home-built legacy systems at Mass General Brigham, and integrating these applications and adapting them to existing workflows has been a challenge. Some functionality was initially disrupted during the go-live, which caught many people off guard, particularly around specimen and slide retrieval workflows, especially for research. These challenges were not necessarily Beaker related; rather, they involved transitioning to other applications and integrating them with Beaker. Many initial integration issues are now resolved, although some workflow optimization is ongoing.
We had a similar pushback as UPMC did regarding going paperless, which, in retrospect, could have been communicated earlier and more clearly. Shortly before go-live, when we emphasized that going digital also meant going paperless, many pathologists felt blindsided and were not onboard. It was a real gripe leading up to the go-live, and even now the transition remains challenging for some, particularly around case organization and tracking in subspecialties where cases are not signed out immediately and require pending IHC or molecular studies. For others, the initial resistance has largely subsided, and the transition has gone well.
Nick Trentadue, there should be considerable satisfaction with Epic’s progress. Its success speaks for itself.
Nick Trentadue (Epic): We are happy with the feedback from pathologists, especially in the past few years. We have a pathology steering group whose members are everyday users of our software and work with us on prioritization, especially with AI development. We’re anticipating delivering three times more code each quarter than we have previously. That’s great, but what do you prioritize? Because when you talk about change management, you can handle only so much. Are we going to focus on digital this quarter? Or outreach tools? Are we going to start bringing in AI summaries and getting rid of the paper case sheet? It’s become a prioritization game.
AI seems to be getting better all the time. I hear this from pathologists who work with it individually and see its application in their departments. Would you encourage every laboratory and pathologist to begin warming up to AI intellectually?
Nick Trentadue (Epic): I don’t see how you can’t. Tasks such as printing the cover sheet or reviewing and gathering data when you open a case can all be fetched, summarized, and available to you. AI is doing that. Faxes, scans, OCR [optical character recognition], whether it’s a report or an order, that’s all structured now using AI models. You’re slowing yourself down and unnecessarily using extra resources if you’re not using it.
Our goal is to use AI to make you more efficient, bring in more cases and specimens, and use your time as a physician more efficiently.
Liron Pantanowitz, I’m sure you agree with Nick, but there’s something else above and beyond for which you’re responsible as chair of an enormous enterprise that needs to harness AI for everyone involved in the enterprise. Tell us about that.
Dr. Pantanowitz (UPMC): I’m pro-AI. We are using traditional AI in routine diagnostic work to take a whole slide image, analyze it, and generate a pathology report. We’re doing it with prostate biopsies, Pap tests, and AFB [acid-fast bacillus] stains, and we’re looking at validating more. We are also using generative AI. Dr. Hooman Rashidi and others in our Computational Pathology and AI Center of Excellence built what we call a Report Genie, and we’ve been using it for a year with bone marrow. It takes the generation of a bone marrow report from 25 to 30 minutes to five or six minutes, and the generative AI along with a rule-based approach writes structured reports the way we like them written.
My problem is that many lab information systems are still transaction based. I would like them to become more like intelligent orchestration platforms. The AI stuff we’re doing is outside the LIS; then we have to figure out how to get it into the LIS, report on it, bill for it, and manage it for compliance and so forth. An example is the digital image management systems. Which way do you go? Is it an image management system that refers out to the LIS to do reporting and ordering and then comes back to the image management system, or vice versa? I’m an LIS-driven person, but we still need to link the LIS to our IMS to get the viewing done on different real estate, like a medical-grade monitor, and then back to the LIS.
Many partnerships are happening, but we are not a true cockpit yet. We’ve found additional middleware plumbing to help us. We use mTuitive to help take our CAP synoptic reports and put them automatically into Epic. The company said it can take an AI output from a different application and put the reports into Epic, in the same way as the synoptic reporting. Is it a workaround or should the LIS itself be doing it? I don’t know, but it would simplify things. We are doing a lot of AI, generative and traditional, but we have to do complicated plumbing to make it seamless. But no matter how amazing or efficient, or that it brings accuracy and standardization, if I make pathologists log in to multiple systems, have to remember passwords, and don’t make it streamlined, they don’t like using it.
I am looking to the future for an intelligent, orchestrated LIS platform that is a diagnostic cockpit and streamlines all of these applications, from difficult flow cytometry analysis and AI prostate analysis to others.
I’m often struck by how much demand and desire is placed on the LIS supplier. Everyone, I think, would endorse what Liron is talking about as a goal, but the difficulties in manpower, programming, and available dollars might stand in the way. Chris Malek, can you comment on that?
Chris Malek (Clinisys): Constraints will always limit what we can do. It will be important, especially when talking about AI, to be able to integrate with other vendors that might be more focused on a specialty, depending on the task. Over time I think we’ll see more embedded features with AI in the LIS that will feel less like a separate tool and more like part of the workflow. That will be critical as we move forward.
Suren Avunjian, can you respond to the demand on LIS suppliers and to what Liron said would be ideal?
Suren Avunjian (LigoLab): This year has changed what software vendors can build. Generative AI has increased our development capacity, but output alone is not the measure; a laboratory needs something reliable in the hands of staff and pathologists. Liron’s point about an intelligent diagnostic cockpit is right, and Chris’ point about integration matters too. No single supplier can build every specialized tool. The LIS should carry case context among systems and make the next action available where the pathologist is working. Otherwise, the “cockpit” becomes a tour of open tabs.
This year we added an MCP [model context protocol] layer to support controlled access to LIS functions and extended Voice beyond dictation to navigation and supported actions. Consider a pathologist reviewing a case in the image system or at the microscope who wants to order a stain or a consult and result the case. The pathologist could state the request, hear which case and specimen the system understood, confirm it, and get an immediate answer if the stain is unavailable or needs a send-out decision. The pathologist stays in control, while the system checks permissions, records the action, and handles the hand-off. Less movement between systems, more time with the case.
Issues are emerging around the use of software and how it will be reimbursed. For example, are we going to penalize the professional for the intelligence and help that software obtains? Omar Baba, does this worry you as you look to the future with the systems you’ll be deploying?
Dr. Baba (Henry Ford): Yes. Reimbursement is always a challenge. We’re eager to deploy AI and the features others have started implementing, but integration is the bottleneck.
It’s fascinating to see how much technology and science have achieved, but now it comes to implementation. This is where the rubber hits the road. Who will pay for it? Especially since digital pathology still isn’t meaningfully reimbursed; the category three CPT codes are contractor priced and rarely paid. We’re tracking those codes. Talks have been given about the business case of going digital, but the return is mostly operational and takes time to realize. When you’re making a case to the C-suite or the hospital’s financial department that you want to adopt a new subscription-based AI tool, what will make your case? How can you justify it? That’s the tricky part.
Joe Sirintrapun, as you look from your perch in Boston, you understand the anxiety professionals have around being penalized for deploying intelligent software and giving up some professional compensation for the efficiency of software. It seems like an unfair paradox, but payment systems are known to have those kinds of built-in problems. Can you comment on that?
Dr. Sirintrapun (Mass General Brigham): Software as a service is turning our professional and technical component billing mindset on its head. There is a fear of losing some of the professional component reimbursement and that AI is taking it away. I had thought it would be part of the technical component, but emerging reimbursement approaches suggest it may be more complicated.
It would be nice if billing were not a zero-sum game, but there are winners and losers. It’s not like the pie necessarily expands. It becomes easier if the pie expands and we’re able to accommodate AI, because in many ways the services we deliver are expanding, yet we’re squeezing a lot of reimbursement into a pie that doesn’t expand.
Also, because AI is integrated in radiology and outside pathology, there’s a lot we can’t control. I’m keeping my ear to the ground about what’s happening on the clinician and radiologist side because those winds are driving much of the discussion. Radiology came from a different angle and has been pursuing new reimbursement models for AI-enabled services, including recognition of software-related costs alongside traditional professional and technical components. We’re going to be living with the consequences of how other fields deal with the issue of how to reimburse for AI.
Nick Trentadue, you must be quite aware of this, particularly as you strategize at Epic. What are your thoughts here?
Nick Trentadue (Epic): A core principle of ours is that before we deploy yet another software model, you must have a good KPI, a good reason for why you’re doing it. The new currency is tokens, and if you can use AI for something, it’s important to know it will impact your reimbursement. If you’re now spending $15 to $20 more on using this [AI] but it cuts your bone marrow report time by 90 percent, that’s worth it. If you’re adding more expense and consuming a lot of tokens, what will ultimately happen? Unless it makes you more efficient and allows you to have fewer staff and sign out more cases, you’re taking away from the same pie.
We’ve been focusing on the financial side of the lab. Groups lose millions of dollars on ABNs [advance beneficiary notice of noncoverage] and on not appealing cases that would be won 98 percent of the time. And payers are using AI too. So from the time the physician places the order, how do we make sure first that it’s appropriate? Does it need an ABN? Is everything that’s outside the lab’s control done? Once it gets to the lab, how can we be as efficient as possible? Once we’re done with our work, how do we ensure we’re rightfully getting paid?
As more of an enterprise vendor, we are deploying AI in all places of the lab, but we also see the touch points and the preanalytical, ordering, and order-to-cash process as necessary to ensure those extra costs can be justified.
In summary, you need to deploy AI but think about it critically—for what model, at what expense, what’s your KPI, what’s your ROI?
Liron Pantanowitz, as pathology chair at UPMC, you’re a virtuoso businessman and you understand the dollars and cents of what you do, because you have many people, systems, and patients who depend on what you’re doing. I’m assuming this discussion is not one you’re hearing for the first time.
Dr. Pantanowitz (UPMC): It isn’t, and I do have to make a lot of business-related decisions. As I want to buy more technology, the question I’m asked by my leadership is, yes, it becomes more expensive and you have to do more with less, so what positions are we eliminating as we use more technology? For me to be able to make business-related decisions, I need data to make data-driven decisions, to justify hiring or to bring on new services or even close services. Most of the data is in the LIS where it’s generated. There are data analytics tools, like Slicer Dicer that Epic provides, but my wish for the LIS is to provide even better data analytics tools, because otherwise it’s the exercise of exporting the data and making dashboards with Power BI and Qlik. That’s extra work, and I don’t know if the data is true data anymore, especially as we try to reconcile with other silos. Additionally, tools like Slicer Dicer are not always optimal for certain pathology workflows.
I’m forever looking at work relative value units, benchmarking them to justify when we should hire, or if I want to hire that the RVUs are growing. A CFO does not just want to look at case volume or turnaround time. They love RVUs.
We’re all saying we should get paid for the work we do, like digital pathology. One of the barriers is we’re not getting paid well enough, and the same goes for AI—will we get paid for doing this? The problem is it’s hard to track what we’re doing. If a slide gets digitized and is available in the LIS as a whole slide image, I would like to track that so I can then generate a code and submit for reimbursement. It’s hard to know if that whole slide image was scanned for the purpose of QA or a research project, or for a tumor board or primary diagnosis. Those are very different reasons, and I can’t bill for nonclinical work.
My wish list is that since the LIS is the treasure box where all the data reside, let’s not move the data out to other systems, but work with us to do better data analytics. We don’t need fancy AI; it’s simple data analytics, and we can use machine learning to improve it, especially if we want predictive functionality. That will add value to the LIS and for LIS users.
Dr. Sirintrapun (Mass General Brigham): I want to add to Liron’s point about orchestration. A good example is the assignment of category three CPT codes for digital pathology. These codes apply to primary diagnostic interpretation, not education or research. For labs that have a hybrid distribution, meaning they distribute glass slides and digital, scalably assigning such category three codes is tricky. To appropriately assign that category three code, you have to know whether the entire case was reviewed digitally. That information is tracked in the digital image management system, yet we don’t have a way to automatically communicate that to the LIS to auto-assign that category three code.
We need more orchestration between the digital IMS and LIS to auto-assign such codes seamlessly. For labs with hybrid distribution and where pathologists can review glass slides or digital, it is not scalable to ask pathologists whether they reviewed all the slides digitally and manually entered the codes in the LIS. There should be an automated way to capture and communicate that information, and the technology to do so exists. Moving forward, I think much of the ground truth in data for digital pathology, like AI result outputs, will reside in digital IMSs, particularly as AI tools are deployed there. There will be ground truth in different systems, and orchestrating that data will be key for our future.
Do financial people in large academic medical centers understand you have an LIS and it now needs a considerable amount of orchestration to answer the very questions they’re asking of you?
Dr. Sirintrapun (Mass General Brigham): They’re not as aware as they should be and, understandably, they have other priorities. Much of what makes orchestration possible is not plug-and-play. Vendors may provide some functionality, but sometimes you have to build it yourself. Either way, there’s an additional cost that often isn’t appreciated. Communicating that need, and justifying the resources to support it, is difficult in the current financial environment.
Omar Baba, what’s your reaction to this bridge that has to be crossed between the financial offices and your own?
Dr. Baba (Henry Ford): It’s a real challenge. We speak in terms of science, patient safety, technology; finance has to weigh the dollars. It’s always a business case to prove—what’s the return for the system? When you implement any new technology, it will take time before you break even, and we’re talking millions of dollars of investment, especially with digital pathology and AI. Building the business case for going digital took time, and this is where everyone is heading. The benefits—remote sign-out, consults, workload balancing, AI readiness—may take several years, and before we get there, there are up-front costs in capital, training, technology, infrastructure. You have to be a visionary and an innovative leader, someone who understands the technology and where things are heading. At the same time, we have to make sure we’re always guarding quality and patient safety. That should be the target in all decisions, not financial return alone.
Nick Trentadue, Epic has a sterling record of speaking to leaders of health care enterprises in an articulate way. You’re a big-ticket item; it requires patience and time on task, but these things are proving to be worthwhile. Do you have tips for your laboratory and pathology customers that you’ve learned over the years?

Nick Trentadue (Epic): We’ve gathered a ton of data the past few years from hospitals that are publicly traded and from Moody’s database, and we’ve seen statistical significance that when organizations use Epic for their clinicals and revenue cycle, they have higher margins. I’m not saying any hospital system has a 30 percent margin, but when you’re looking at 0.8 or 1.2 percent versus 3 or 3.24 percent, the groups that use Epic’s clinical and revenue cycle systems have statistically higher margins. Not everything is causation, but that is what we track.
We feel strongly about having both sides of the house integrated, where you have data and can see where your ABNs and denials are higher and where you’re doing unnecessary testing. Technology can solve for those things. We’ve seen that time and time again, and we’ll continue to ensure the order-to-cash process is as efficient as possible.
I’ve been asking pathologists lately about how much they want to talk to their computer or their computer to talk to them, and I have seen an interesting bell curve of responses. When we develop AI tools in the future, we have to look at it from different angles. Will you be redoing old-school transcription; you just don’t need the transcriptionist? Or are we interacting with you in a way that there might be confirmations and a click? Or are you asking a question in a generative AI sense? We’re trying to work with pathologists to get that right, knowing it’s not going to be the same for everyone.
Chris Malek, can you give us your closing thoughts?
Chris Malek (Clinisys): AI and data governance, auditability, and transparency will be a critical piece. We and many vendors are working on making sure this is transparent and you can see how the AI assisted, what decisions it made, what decisions a human made, and keeping a human in the loop to prove that.
Human in the loop is an important concept that can’t be ignored. Suren Avunjian, a final thought from you?
Suren Avunjian (LigoLab): Chris is right to emphasize governance and auditability. The human in the loop should be a safety net, not a second shift created by the software. In AI-assisted requisition processing, for example, we need to know how often the system extracts the right information, what staff must verify, how many cases it sends to exception review, and how long corrections take. If the review consumes the time we claimed to save, the workflow needs work. We also need a record of what the system did and what a person approved or changed, especially when that decision affects a case or a bill.
I agree with Liron that laboratories need insights they can act on, not another mountain of data. A manager should be able to set meaningful thresholds and see, for example, where cases are aging, orders lack information, or denials are beginning to cluster. The alert has value only if it reaches someone who can resolve the issue while the outcome can still change.
That gives vendors a practical standard: Are pathologists able to stay focused on the case? Do staff see and clear exceptions sooner? Does case information stay consistent across the LIS, imaging, and financial workflow? Can the laboratory show better turnaround, quality, and financial performance after accounting for the work added by review? AI is an opportunity to improve those steps. We should be willing to show the work removed, the oversight retained, and the results over time.
Liron Pantanowitz, what should people in the field have top of mind as we look toward the end of this year and into the next?
Dr. Pantanowitz (UPMC): Most people do not understand the story of the lab. That’s bad if you’re the CFO or COO and you have to make decisions, especially those that impact the lab, for resources, staffing, et cetera. It’s on us to tell a better story. When you tell all the stories, what comes out is that the lab is the diagnostic ecosystem for the whole enterprise. If the lab is the diagnostic ecosystem, the LIS is the CNS, the central nervous system. It’s the digital backbone to everything. It connects patients to their specimens, the data from the preanalytical to analytical and postanalytical, and all the compliance data. It’s a pity we have to take all that data and move it to some other data warehouse. We do it at UPMC, along with 70 other IT systems. I sent Dr. Matthew Hanna to work with the data warehouse, and they don’t know what to do with all the data. They don’t understand it, and they don’t know how to manipulate it in dashboards.
The LIS has a great opportunity to support not just the lab but an entire ecosystem, whether you’re a reference lab, a small private practice, or a large integrated health care system like ours. I want to see us move away from being transaction based to more of an orchestrated platform so we can do our diagnostic work there and not have to keep moving out into complicated applications, because the message is that CIOs want simplicity. They want one system for everyone to work in. If we can’t do our work there, I have to look at complicated plumbing. Over time, LIS vendors have to make a decision: How much do they want to own and how much do they want to incorporate from an image management system and generative AI and so forth?
I also want all the data analytics to be available so I can run my dashboards. The cherry on top is whether you can help me do the reconciliation with the finance systems. Is someone doing the correct billing? Are they getting reimbursed? Are we doing the right collections? That would avoid a complicated setup for me, especially as someone who has to oversee the governance of the laboratory.