Harrison.ai

In 2015, a medical student named Aengus Tran took a two-month clinical placement at a short-staffed cardiology hospital in Ho Chi Minh City. He had planned to specialise in cardiology. Instead he watched patients wait for diagnoses that arrived too slowly, held up not by a lack of medicine but by a lack of hands to read the results.

Three years later, in 2018, he and his brother Dimitry Tran founded Harrison.ai in Sydney. The premise was narrow and practical: much of modern diagnosis is pattern recognition on images - X-rays, CT scans, pathology slides - and there are never enough trained specialists to keep pace with rising volumes. Software that reads those images could give clinicians back the one thing they cannot manufacture: time.

Harrison.ai's answer is not a robot doctor. Its radiology system, sold under the brand annalise.ai, sits inside the hospital workflow and does something deliberately modest - it looks at every incoming scan, flags the ones that cannot wait, and drafts a structured report for the radiologist to check. The specialist stays in charge. The queue simply gets smarter.

That framing - amplify the clinician, do not replace them - runs through everything the company builds, from its pathology joint venture to a radiology-specific language model that can describe what it sees in an image and write it up.

The problem it solves. Radiology and pathology volumes have grown faster than the workforce trained to interpret them. A radiologist may read thousands of images a week, most of them normal. Buried in that stream is the occasional scan showing a bleed, a fracture or an early sign of stroke - the case where minutes matter. Harrison.ai's tools scan everything and surface the urgent findings, so critical patients move up the list and reports get drafted automatically.

Who uses it. Its software is live at more than 1,000 healthcare sites across four continents, with around 96 healthcare organisations as customers and over 3,400 clinicians using the tools. In radiology, annalise.ai supports the care of more than 6 million patients a year. The United States is already the company's third-largest revenue market, and the stated aim is to make it the largest.

The expertise behind it. The company is deliberately clinician-led - built by practising doctors alongside engineers. Co-founder and CEO Aengus Tran trained as a doctor at UNSW and is also a world-ranked AI engineer and data scientist, a dual background the company leans on when designing tools that have to earn the trust of specialists.

Harrison.ai is a B2B business. Its products are licensed to health systems as recurring subscriptions that plug into existing radiology and pathology workflows. The company says its contracted annual recurring revenue tripled for three consecutive years after it began monetising annalise.ai in 2022, and in the US it has secured a Medicare reimbursement pathway (New Technology Add-on Payment) for one of its CT brain algorithms - a signal that the tools are being paid for as clinical infrastructure, not pilots.

How it differs. Two choices stand out. First, it is clinician-led by design - doctors help build and validate the models, which matters in a field where specialists must trust the output. Second, rather than walling off its technology, it runs an Open Platform that hosts competitors' models alongside its own, betting that being the integration layer is worth more than being the only model in the room.