The clinical inbox is where small tasks go to become a second shift. A refill request looks harmless until a physician has to find the last visit, medication history, diagnosis and recent labs. A normal result looks easy until it needs to be reviewed, documented and explained to a patient. A specialist's note is only one document - sometimes dozens of pages long. Multiply those jobs across a full panel, and the inbox follows clinicians home.
Droxi built a business around that unglamorous pileup. The Boston-based company puts AI agents into existing electronic health record workflows, where they prepare or complete repetitive administrative work. It does not present itself as a synthetic doctor. It presents itself as the colleague who has already opened every relevant drawer before the physician sits down.
That distinction matters. In healthcare, an impressive answer is not enough; a user needs to know where it came from, what the software changed and who remains accountable. Droxi's product design keeps the clinician at the decision point. The machine gathers context, drafts, flags and, where configured, closes routine work. The care team decides.
The first idea was a symptom
Gadi Shenhar's motivation was personal. A medication error harmed his grandmother. He and co-founder Raveh Ben Simon initially approached the problem as one of preventing errors in treatment. Then they did the most useful thing a founder can do with a powerful origin story: they let customers argue with it.
After speaking with hundreds of doctors, the pair concluded that errors were often the visible consequence of a broader operating problem. Medical staff were overloaded by administrative work, navigating menus and codes while trying to preserve enough attention for patients. That changed the product thesis. Instead of building a general error-prevention system, Droxi attacked the recurring environment that makes distraction and fatigue ordinary.
The founders had a useful combination of technical credibility and exposure to clinical software. Shenhar had been chief data officer at medical-imaging AI company Aidoc. Ben Simon had worked as a software engineer at Qualcomm. Both studied electrical engineering at the Technion and served in Israel's Unit 8200. They founded Droxi in 2022. By December 2024, the company had raised a $21 million Series A led by Drive Capital, with Verissimo Ventures, AirAngels, Lurie Consulting and Elad Litman participating. Contemporary reporting put total funding at $25 million. No valuation was disclosed.
The system doesn't tell doctors what to do; it simply makes the information accessible.Gadi Shenhar, co-founder and CEO
What the machine actually does
Droxi now packages its platform as two products. Unbox is the clinical-inbox layer. For lab results, it can automatically close normal results according to configured rules and send a customizable patient message. Abnormal or complex results arrive with a provider summary and patient-friendly draft, with recommendations traceable to source data. For prescription refills, it consolidates requests from fax, portal and e-prescription channels, then surfaces medication history, recent labs, diagnoses and duplicate requests in one view. For consult notes and other clinical documents, it extracts diagnoses, treatment plans and medication changes, linking summary points back to the original text.
The product loop / simplified
Unlock is the newer expansion. It tackles Advanced Primary Care Management, a Medicare program launched in 2025 that pays practices a monthly amount for managing eligible patients between visits. Droxi says Unlock reads chart data to flag eligible patients, manages outreach and consent, drafts editable care plans, supports billing documentation and helps maintain patient engagement. Every plan and billing item remains subject to practice review. The company lists 2026 national payment rates of roughly $16, $54 or $116 per enrolled patient per month, depending on complexity and locality.
The pairing is shrewd. Unbox sells hours back; Unlock sells a path to recurring revenue. One attacks a visible cost, the other helps monetize work practices should already be doing. Both use the same distribution advantage: they live with the chart data and the workflow rather than asking clinicians to adopt another disconnected destination.
These are Droxi's company-reported deployment figures, not independently audited benchmarks.
Who buys - and what it costs
Droxi sells business-to-business software to primary-care practices, community health centers, rural systems and provider groups. Physicians, nurse practitioners and physician assistants use the clinical workflow; practice managers and operations leaders buy around burnout, staffing capacity, response time and revenue. Publicly named customers include Blue Ridge Health, Salience Health, Northern Counties Health Care, Rainbow Pediatric, Stedman-Wade Health Services and Primary Care Partners of South Bend.
Public pricing is not posted. Prospective buyers are sent to a sales call, which signals an enterprise subscription shaped by provider count, inbox volume, integration and implementation. The relevant cost comparison is broader than a software seat. A practice must weigh the contract against clinician hours, overtime, unfilled capacity, staff turnover and, for Unlock, APCM revenue that otherwise goes uncaptured. Without Droxi's actual quote, any sharper return-on-investment claim would be theatre.
There is customer evidence, although much of it is presented by the company. Salience Health physician Jason Conn says Droxi saves him about an hour a day. Blue Ridge Health clinicians describe accurate summaries and less digging. Droxi reports that one Blue Ridge deployment saved 46 hours in five weeks and projected 215 hours returned to the clinic monthly at full adoption. Those results are useful proof points, but buyers should reproduce the baseline with their own inbox categories before signing.
The trust is the moat
Droxi competes with the status quo first: clinicians doing the work manually, staff triage, and whatever native tools an EHR already supplies. It also sits beside focused inbox competitors such as Affineon and a growing crowd of ambient-AI companies moving beyond visit notes. Its differentiation is less about a magical model than about workflow placement. The software brings patient context to the task, supports provider-specific rules, traces summaries to source material and avoids demanding a new screen.
That is also why athenahealth matters. Droxi is listed in its marketplace and has participated in the vendor's builder ecosystem. Unlock supports athenahealth practices today; eClinicalWorks support is described as in progress. Integrations with large EHR platforms can become distribution, product infrastructure and credibility at once. They can also become dependency. If access changes, interfaces break or a native vendor ships a comparable feature, the startup has to prove that its workflow depth is worth keeping.
There is a second competitive line hiding inside the interface: explainability. A generic assistant can produce a plausible paragraph, but a clinician needs to jump from a claim to the original note or result. Droxi says its document summaries link each point to source text and its recommendations remain traceable. That feature sounds modest until the software is wrong. Then provenance is the difference between a quick correction and a scavenger hunt. In this category, the recovery path is part of the user experience.
Where the playbook stops working
- An unsupported EHR turns a workflow product into another tab.
- Bad chart data produces neat summaries of unreliable material.
- Low inbox volume weakens the time-savings case.
- Practices unwilling to configure rules will capture less safe automation.
- Clinical, security or compliance policies may prohibit required access and actions.
What builders can steal
The copyable move is not “add AI to healthcare.” It is the sequence. Start with a painful event. Interview until the anecdote gives way to a repeated system. Choose a workflow with measurable friction. Enter the software people already use. Automate the clerical layer before touching judgment. Show the source. Measure minutes returned, not tokens generated.
Droxi's own shift makes that lesson unusually clean. The grandmother's medication error gave Shenhar a reason to care. Hundreds of physicians supplied the reason to change direction. The first concept did not fail as a mission; it failed as the best product boundary. Administrative overload was broader, more frequent and easier to observe. The company could sell a concrete before-and-after: fewer screens, fewer clicks, fewer routine items waiting at night.
The next chapter will test whether that narrow credibility stretches. A company known for clearing inboxes is now helping practices operate a reimbursed care program. If Unlock works, Droxi becomes more than a productivity layer; it becomes operational infrastructure tied to practice economics. If it does not, Unbox remains a useful but contested wedge in a market where EHR companies and AI vendors both want the same screen.
For now, Droxi's most persuasive idea is also its least theatrical: the doctor should see fewer things, with better context, and go home sooner. In an industry full of systems asking clinicians for attention, a product that gives attention back has a straightforward reason to exist.