The modern office has a peculiar talent for turning essential work into an inbox. In healthcare, the transformation is especially complete. A refill request waits beside a lab result. A scanned document lands near a patient message. Each item looks small until someone must hunt through an electronic record, gather the context, decide what matters, and close the loop. By evening, the queue has behaved like a hydra with a login screen.
Gadi Shenhar built a company around that queue. Droxi, which he co-founded with Raveh Ben Simon in 2022, works inside electronic health record systems and helps clinical teams process routine inbox tasks. The software reads the surrounding record, brings relevant information together, automates work a clinician has configured, and elevates what requires human review. The pitch is deliberately narrower than the grand promises attached to artificial intelligence. Shenhar wants the machine to absorb the repetition, not impersonate the professional.
That boundary is the interesting part. Shenhar is an engineer by training, but his public argument is less about what a model can do than about where a person’s attention belongs. “We’re not in a day where AI can replace anybody,” he said while introducing Droxi to the American market in 2023. For a founder selling automation, restraint is an unusual calling card. It is also a practical one. In healthcare, confidence must be earned one workflow at a time.
The first idea lost the interview
Droxi did not begin with the clinical inbox as its fixed destination. Shenhar has said that he and Ben Simon initially thought about medication error. Then they did something that sounds obvious and often proves difficult: they let customer conversations overrule the founding idea. They spoke with hundreds of physicians and asked them to rate the pain points in their work. Again and again, the administrative labor behind care received the emphatic score.
Nobody needed a survey to learn that doctors dislike paperwork. The interviews revealed something more useful: the work behind the scenes had become a coherent, urgent product problem. The inbox was where fragmented information, repetitive action, and scarce judgment collided. What looked like a list of chores was really a map of hidden labor.
“Our company helps physicians and medical staff spend more time on patients and less time on paperwork.”Gadi Shenhar
There is a founder lesson here, and it has nothing to do with medicine. Customer discovery is useful only if the customer is allowed to ruin your favorite slide. Shenhar’s interviews did not decorate the original thesis. They replaced it with a sharper one. The company moved toward the stubborn queue everyone had learned to tolerate.
A career spent behind the interface
Shenhar studied electrical engineering at the Technion from 2017 to 2021, focusing on areas that included machine learning, biological systems, and computers. He also co-authored a 2020 machine-learning preprint on predictive pooling. From the start, his technical path kept returning to data under pressure: how to collect it, interpret it, and turn it into action.
Before Droxi, Shenhar was the third employee at Aidoc, according to his public founder biography. He went on to lead data work there as the company’s radiology AI reached hospitals. He also worked as a software engineer at Speedata, dealing with algorithms and data-intensive systems. Aidoc supplied the clinical environment. Speedata supplied another view of computational scale. Droxi combined the lessons around a less glamorous object: the everyday workflow.
Electronic records are full of facts, but facts do not arrive in narrative order. A refill request may depend on a recent result, another prescription, a past note, or an instruction hiding elsewhere in the chart. The clinical worker supplies the missing continuity, often by clicking from window to window. Droxi’s product thesis is that software can assemble much of that context before asking for attention.
This explains Shenhar’s emphasis on transparency and explainability. The product must show enough of its work to be trusted, and it must live inside the systems already used by clinics. A clever answer in a separate tab is still another tab. Workflow software succeeds when it removes detours.
The arithmetic of an evening returned
Droxi’s reported deployment figures are concrete: more than five hours returned to each provider per week, 38 percent of routine inbox work handled automatically, and a 52 percent reduction in after-hours clinical inbox time. These are company-reported numbers, not a universal guarantee. Their useful unit is time rather than novelty.
That focus gives Droxi a legible product story. The work arrives as dozens of small interruptions, so the benefit must be recomposed into time. Five hours translates an abstract efficiency score into a Friday afternoon, a week of unrushed lunches, or a stack of tasks that does not follow someone home.
The product still needs a clinician’s choices. Teams configure what can be handled routinely; the system summarizes and surfaces what deserves review. Shenhar frames automation as backup rather than a coup. This makes the product’s stopping point as important as its capability. “Don’t accept today’s reality in healthcare as it is,” he wrote in 2026. The alternative he described was technology carrying administrative weight while doctors practiced medicine.
Haifa code, Ohio introductions, Boston ambition
Building the product was only one translation problem. Selling it in the United States required another. Droxi began with an Israeli team and joined the Youngstown Business Incubator in Ohio in 2023. Shenhar spoke candidly then about the difficulty of learning a new market’s culture and language. The incubator supplied introductions to regional healthcare providers and a practical bridge into the country.
At the time, Droxi employed ten people and described early US work in California and Montana. The scale changed quickly. In December 2024, the company announced a $21 million Series A led by Drive Capital, with participation from Verissimo Ventures, AirAngels, Lurie Consulting, and Elad Litman. Shenhar said the round brought total funding to more than $25 million in two years. The announcement thanked the team by name, an operator’s inventory tucked inside the celebration.
By December 2025, Shenhar announced a Boston office and began recruiting what he called the founding Boston team. The geography traces a company learning its market: Haifa for engineering roots, Youngstown for a first American network, Boston for a deeper bench of clinical and technology talent. Each city answers a different question.
Shenhar’s recruiting voice offers a glimpse of the person behind the role. In one Ohio hiring post, he promised prospective candidates that he was nice and would respond, then admitted that his messages sometimes became “slightly overflooded” like the physicians his company served. It was a small joke with suspiciously good product positioning. Some founders carry an elevator pitch. Shenhar appears to carry an inbox metaphor.
Ambition, with a stop button
Shenhar’s aspiration is expansive even when his product language is careful. In May 2026, he described returning time, attention, and energy to clinicians as “the dream of my life.” The phrase is grander than his usual operational vocabulary, which is precisely why it lands. Beneath the integrations, deployment metrics, and funding announcements sits a very personal allocation problem: who gets to spend human attention, and on what?
His answer has remained consistent. Clinicians should make the decisions that require clinical judgment. Software should prepare context, handle configured routine work, and quiet the noise. It is a useful stance in an industry crowded with declarations that every new model will reinvent everything by Tuesday.
“This is the dream of my life. This is our purpose and why we wake up in the morning and go to work.”Gadi Shenhar
There are harder chapters ahead. Healthcare integrations are slow. Trust is fragile. Every new customer brings its own habits, systems, and thresholds for autonomy. A larger company must keep listening with the intensity of a small one. The interviews that found the clinical inbox cannot become a charming origin story; they have to remain an operating method.
For now, Shenhar’s work can be understood through one tidy reversal. The clinical inbox used to be the place where other work accumulated. He made it the work itself. Then he asked software to clear what it could, explain what it did, and leave the consequential choices where they belong. There are louder visions for AI. Few are as easy to picture at 6:03 p.m., when the last small task finally disappears.