Most discussions about AI in pathology begin with what an algorithm can detect. In the laboratory, a more practical question often determines whether it gets used at all: where does the result end up, and what does the pathologist need to do to reach it? 

The answer often involves a system that is central to everyday laboratory work: the laboratory information system (LIS), which manages case information and connects it with other parts of the diagnostic process. 

AI tools that analyze tissue images and flag areas that may need closer examination are starting to move from research into clinical use. The quality of the analysis matters, of course. But whether a tool is used day to day depends just as much on how its results fit into the laboratory’s work. 

AI needs a place in the workflow 

A pathology workflow is much more than a pathologist looking at a slide. A case passes through registration, preparation, scanning, interpretation, and reporting, and at almost every step someone needs information that lives in another system. 

Consider an AI application that highlights regions of interest, areas of a digital tissue image that may warrant closer examination. The analysis may help the pathologist focus on the most relevant parts of the slide. But if the result sits in a separate application, the pathologist has to jump between screens to check it against the sample details, earlier findings, and the rest of the case. The tool helps, but it also adds work. 

Much of this comes down to interoperability. A typical pathology lab already uses an LIS, digital pathology software, scanners, image management systems, and other hospital systems. An AI tool has to exchange data with the right ones without losing track of which patient, sample, and image a result belongs to. 

And the connections don’t stop at pathology. In cancer diagnostics, the same specimen is increasingly analyzed by several laboratory specialties, so sample workflows and information systems have to work across department lines. According to Päivi Elo, Senior Product Expert at Mylab, this requires close collaboration between laboratories, “supported either by the capabilities of a shared laboratory information system or by increasingly seamless integration between multiple information systems.” 

New AI tools will keep arriving. The challenge is to make them work with the systems and workflows the lab already uses, rather than adding another isolated application. So when you evaluate one, look past the accuracy figures. Ask where its results will show up, who will see them, and at what point in the diagnostic process. 

Päivi also argues that the scope should be wider than diagnostic review:  

“To maximize the value of artificial intelligence in pathology laboratories, AI must have access to and be able to utilize information across the entire pathology system. Limiting AI to supporting pathologists’ diagnostic review will not be sufficient; it must also be leveraged to support laboratory operations and workflows.” 

AI could, for example, help laboratories prioritize incoming cases or plan how work is distributed across the team. These uses are different from analyzing tissue images, but they still depend on access to the right information at the right time. 

Good AI depends on good data 

Pathology generates sample records, images, and diagnostic findings, along with a growing share of molecular data. Much of it stays relevant for years after the original examination. 

When a new case comes in, an earlier biopsy or a genetics result from another lab can change how it is read. As pathology works more closely with genetics and molecular diagnostics, access to information across disciplines becomes more important. That gets difficult when the information sits in separate systems or in formats that don’t match. 

The European Health Data Space (EHDS) Regulation, which entered into force in 2025, will introduce new requirements for sharing and using electronic health data across the EU. For laboratories, it adds another reason to pay attention to how diagnostic information is structured, stored, and made accessible. 

The same problem shows up when pathology data is used for AI. Historical data collected over many years can be valuable for research, biomarker development, and the training and evaluation of AI models. That holds only if the data is consistent and traceable, and the right permissions and safeguards are in place. 

Archives matter for the individual patient too. As Päivi points out, historical records and preserved specimens serve as a reference when new samples are analyzed. As diagnostic methods advance, it may also become possible to re-evaluate earlier specimens and refine a previous interpretation. 

Volume alone doesn’t help. Ten years of results recorded three different ways can mean a lot of cleanup before the data is useful. 

The LIS holds an important part of the laboratory’s diagnostic record. How that information is structured and maintained affects what it can be used for later, especially when a lab moves to a new system. 

The pathologist remains at the center 

AI can take on specific tasks. The diagnosis is still the pathologist’s. That means the pathologist needs to see what a tool has produced, check it, and weigh it against everything else known about the case. 

Validation, documentation, and human oversight are part of the deal. They are easiest to get right when they are planned into the workflow from the start, rather than checked off during procurement. 

Making room for what’s next

Pathology will keep changing, with new methods, new imaging, and new uses for laboratory data. Each of them puts its own demands on the systems people work in every day. 

No one can predict exactly which tools will matter in five years. What laboratories can influence is whether their environment can take in new capabilities without being rebuilt each time. 

In Päivi’s view, this calls for standardized connectivity and open interfaces, so that new integrations can be deployed “more efficiently and with minimal custom development effort.” 

She also sees a shift in how work is organized, from sample-centric processing toward user-specific worklists that help laboratory professionals organize and prioritize their daily tasks. For laboratories facing growing case volumes and limited specialist capacity, that kind of flexibility can matter as much as any single new tool. 

So before choosing an AI tool, a laboratory can learn a lot from looking at its own environment first. Does information move between systems without someone retyping it? Do results show up where people actually work? And will today’s records still make sense after the next system change? 

None of these are really AI questions. But answering them is what it takes to make technology, workflows, and data work together. 

A foundation for connected pathology

Discover how My+ Pathology supports connected diagnostic workflows, structured data, and information exchange across systems.