Elea AI is chasing the healthcare productivity opportunity by targeting pathology labs’ legacy systems

VC funding into AI tools for healthcare was projected to hit $11 billion last year — a headline figure that speaks to the widespread conviction that artificial intelligence will prove transformative in a critical sector.

Many startups applying AI in healthcare are seeking to drive efficiencies by automating some of the administration that orbits and enables patient care. Hamburg-based Elea broadly fits this mould, but it’s starting with a relatively overlooked and underserved niche — pathology labs, whose work entails analyzing patient samples for disease — from where it believes it’ll be able to scale the voice-based, AI agent-powered workflow system it’s developed to boost labs’ productivity to achieve global impact. Including by transplanting its workflow-focused approach to accelerating the output of other healthcare departments, too.

Elea’s initial AI tool is designed to overhaul how clinicians and other lab staff work. It’s a complete replacement for legacy information systems and other set ways of working (such as using Microsoft Office for typing reports) — shifting the workflow to an “AI operating system” which deploys speech-to-text transcription and other forms of automation to “substantially” shrink the time it takes them to output a diagnosis.

After around half a year operating with its first users, Elea says its system has been able to cut the time it takes the lab to produce around half their reports down to just two days.

Step-by-step automation

The step-by-step, often manual workflow of pathology labs means there’s good scope to boost productivity by applying AI, says Elea’s CEO and co-founder Dr. Christoph Schröder. “We basically turn this all around — and all of the steps are much more automated … [Doctors] speak to Elea, the MTAs [medical technical assistants] speak to Elea, tell them what they see, what they want to do with it,” he explains.

“Elea is the agent, performs all the tasks in the system and prints things — prepares the slides, for example, the staining and all those things — so that [tasks] go much, much quicker, much, much smoother.”

“It doesn’t really augment anything, it replaces the entire infrastructure,” he adds of the cloud-based software they want to replace the lab’s legacy systems and their more siloed ways of working, using discrete apps to carry out different tasks. The idea for the AI OS is to be able to orchestrate everything.

The startup is building on various Large Language Models (LLMs) through fine-tuning with specialist information and data to enable core capabilities in the pathology lab context. The platform bakes in speech-to-text to transcribe staff voice notes — and also “text-to-structure”; meaning the system can turn these transcribed voice notes into active direction that powers the AI agent’s actions, which can include sending instructions to lab kit to keep the workflow ticking along.

Elea does also plan to develop its own foundational model for slide image analysis, per Schröder, as it pushes towards developing diagnostic capabilities, too. But for now, it’s focused on scaling its initial offering.

The startup’s pitch to labs suggests that what could take them two to three weeks using conventional processes can be achieved in a matter of hours or days as the integrated system is able to stack up and compound productivity gains by supplanting things like the tedious back-and-forth that can surround manual typing up of reports, where human error and other workflow quirks can inject a lot of friction.

The system can be accessed by lab staff through an iPad app, Mac app, or web app — offering a variety of touch-points to suit the different types of users.

The business was founded in early 2024 and launched with its first lab in October having spent some time in stealth working on their idea in 2023, per Schröder, who has a background in applying AI for autonomous driving projects at Bosch, Luminar and Mercedes.

Another co-founder, Dr. Sebastian Casu — the startup’s CMO — brings a clinical background, having spent more than a decade working in intensive care, anaesthesiology, and across emergency departments, as well as previously being a medical director for a large hospital chain.

So far, Elea has inked a partnership with a major German hospital group (it’s not disclosing which one as yet) that it says processes some 70,000 cases annually. So the system has hundreds of users so far.

More customers are slated to launch “soon” — and Schröder also says it’s looking at international expansion, with a particular eye on entering the U.S. market.

Seed backing

The startup is disclosing for the first time a €4 million seed it raised last year — led by Fly Ventures and Giant Ventures — that’s been used to build out its engineering team and get the product into the hands of the first labs.

This figure is a pretty small sum vs. the aforementioned billions in funding that are now flying around the space annually. But Schröder argues AI startups don’t need armies of engineers and hundreds of millions to succeed — it’s more a case of applying the resources you have smartly, he suggests. And in this healthcare context, that means taking a department-focused approach and maturing the target use-case before moving on to the next application area.

Still, at the same time, he confirms the team will be looking to raise a (larger) Series A round — likely this summer — saying Elea will be shifting gear into actively marketing to get more labs buying in, rather than relying on the word-of-mouth approach they started with.

Discussing their approach vs. the competitive landscape for AI solutions in healthcare, he tells us: “I think the big difference is it’s a spot solution versus vertically integrated.”

“A lot of the tools that you see are add-ons on top of existing systems [such as EHR systems] … It’s something that [users] need to do on top of another tool, another UI, something else that people that don’t really want to work with digital hardware have to do, and so it’s difficult, and it definitely limits the potential,” he goes on.

“What we built instead is we actually integrated it deeply into our own laboratory information system — or we call it pathology operating system — which ultimately means that the user doesn’t even have to use a different UI, doesn’t have to use a different tool. And it just speaks with Elea, says what it sees, says what it wants to do, and says what Elea is supposed to do in the system.”

“You also don’t need gazillions of engineers anymore — you need a dozen, two dozen really, really good ones,” he also argues. “We have two dozen engineers, roughly, on the team … and they can get done amazing things.”

“The fastest growing companies that you see these days, they don’t have hundreds of engineers — they have one, two dozen experts, and those guys can build amazing things. And that’s the philosophy that we have as well, and that’s why we don’t really need to raise — at least initially — hundreds of millions,” he adds.

“It is definitely a paradigm shift … in how you build companies.”

Scaling a workflow mindset

Choosing to start with pathology labs was a strategic choice for Elea as not only is the addressable market worth multiple billions of dollars, per Schröder, but he couches the pathology space as “extremely global” — with global lab companies and suppliers amping up scalability for its software as a service play — especially compared to the more fragmented situation around supplying hospitals.

“For us, it’s super interesting because you can build one application and actually scale already with that — from Germany to the U.K., the U.S.,” he suggests. “Everyone is thinking the same, acting the same, having the same workflow. And if you solve it in German, the great thing with the current LLMs, then you solve it also in English [and other languages like Spanish] … So it opens up a lot of different opportunities.”

He also lauds pathology labs as “one of the fastest growing areas in medicine” — pointing out that developments in medical science, such as the rise in molecular pathology and DNA sequencing, are creating demand for more types of analysis, and for a greater frequency of analyses. All of which means more work for labs — and more pressure on labs to be more productive.

Once Elea has matured the lab use case, he says they may look to move into areas where AI is more typically being applied in healthcare — such as supporting hospital doctors to capture patient interactions — but any other applications they develop would also have a tight focus on workflow.

“What we want to bring is this workflow mindset, where everything is treated like a workflow task, and at the end, there is a report — and that report needs to be sent out,” he says — adding that in a hospital context they wouldn’t want to get into diagnostics but would “really focus on operationalizing the workflow.”

Image processing is another area Elea is interested in other future healthcare applications — such as speeding up data analysis for radiology.

Challenges

What about accuracy? Healthcare is a very sensitive use case so any errors in these AI transcriptions — say, related to a biopsy that’s checking for cancerous tissue — could lead to serious consequences if there’s a mismatch between what a human doctor says and what the Elea hears and reports back to other decision makers in the patient care chain.

Currently, Schröder says they’re evaluating accuracy by looking at things like how many characters users change in reports the AI serves up. At present, he says there are between 5% to 10% of cases where some manual interactions are made to these automated reports which might indicate an error. (Though he also suggests doctors may need to make changes for other reasons — but say they are working to “drive down” the percentage where manual interventions happen.)

Ultimately, he argues, the buck stops with the doctors and other staff who are asked to review and approve the AI outputs — suggesting Elea’s workflow is not really any different from the legacy processes that it’s been designed to supplant (where, for example, a doctor’s voice note would be typed up by a human and such transcriptions could also contain errors — whereas now “it’s just that the initial creation is done by Elea AI, not by a typist”).

Automation can lead to a higher throughput volume, though, which could be pressure on such checks as human staff have to deal with potentially a lot more data and reports to review than they used to.

On this, Schröder agrees there could be risks. But he says they have built in a “safety net” feature where the AI can try to spot potential issues — using prompts to encourage the doctor to look again. “We call it a second pair of eyes,” he notes, adding: “Where we evaluate previous findings reports with what [the doctor] said right now and give him comments and suggestions.”

Patient confidentiality may be another concern attached to agentic AI that relies on cloud-based processing (as Elea does), rather than data remaining on-premise and under the lab’s control. On this, Schröder claims the startup has solved for “data privacy” concerns by separating patient identities from diagnostic outputs — so it’s basically relying on pseudonymization for data protection compliance.

“It’s always anonymous along the way — every step just does one thing — and we combine the data on the device where the doctor sees them,” he says. “So we have basically pseudo IDs that we use in all of our processing steps — that are temporary, that are deleted afterward — but for the time when the doctor looks at the patient, they are being combined on the device for him.”

“We work with servers in Europe, ensure that everything is data privacy compliant,” he also tells us. “Our lead customer is a publicly owned hospital chain — called critical infrastructure in Germany. We needed to ensure that, from a data privacy point of view, everything is secure. And they have given us the thumbs up.”

“Ultimately, we probably overachieved what needs to be done. But it’s, you know, always better to be on the safe side — especially if you handle medical data.”