Natalia Neha Khosla

The through-line

Almost everything Natalia Neha Khosla says circles back to a single claim: the people standing at the point of care are the best-positioned people to identify what healthcare technology should actually do, and they are systematically absent from the rooms where that technology gets decided [1][3]. She frames this from her own path, keeping notes from rotations at the University of Chicago on a clipboard, ideas she now recognises as "point of care problem identification and uh leading to potential solutions" [1], and then building those ideas into Simbie AI's voice agents. The argument has an edge of alarm to it. She reports being "in rooms where there are people deciding technology that's going to be built and there are no physicians in the room. And this is healthcare technology where the end user is physicians and patients" [1], and she treats that absence as the direct cause of the tone-deaf tools clinicians already live with.

Her second, tightly linked preoccupation is methodological: the problem is durable, the solution is disposable. She wants clinicians to understand that the knowledge they already hold is the scarce asset, and that the technical work is learnable afterwards [1].

On why physicians have to be in the room

The corporatisation of medicine and the disappearance of clinicians from leadership are, in her account, the same trend viewed from two angles. She cites the statistic that "today, fewer than 5 % of CEOs of major health systems, for example, are physicians", calls it "jarring" and "pretty disturbing", and notes that as that number declined, corporatisation of the field moved "almost mirroring that exactly" [1]. She is careful not to make it a character attack: "I really don't mean this to say anything against non-physician uh or non-clinician leaders. Like I think that the people doing this work are very well-intentioned they really care about patients and about outcomes" [1]. The deficit is experiential rather than moral. A physician has spent "thousands of hours in a room with a patient holding their hand", knowing "their story, you know where they live, what they eat, you know their bowel habits, you know their sister, their mother", a level of understanding "you can never match with a data sheet, which is what patients become when you're looking in board rooms" [1].

Her sharpest formulation of the cost is the metric that never gets counted. A profit and loss sheet can show that money was saved or throughput increased, "But the invisible metric that never gets found is, okay, well you just spurred provider burnout in 20 % of the department and now people are planning to retire early" [1]. The same blindness shows up in product decisions: if you optimise patient satisfaction or length of stay as numbers, you never see "the patient's emotions, their interaction with the staff, their interaction with their family that are causing them to need to stay an extra day" [1]. She ties this directly to physician advocacy, which she treats as essential to addressing burnout and working conditions rather than as a separate concern [1][3].

On staying in the problem space

Khosla's core startup discipline is that the problem deserves nearly all of the founder's attention. In agile or iterative methodology, she argues, "the problem space is the most important space. The solution will change, should be expected to change" [1]. She takes the pace of the field as proof: "There's literally a new voice AI model coming out essentially every four days right now, so we have to be agile" [1]. What endures is understanding, and she notes that companies "spend millions and millions and billions of dollars to find people who actually understand the point of care problem" [1], value she did not recognise she was accumulating as a student.

The inversion of this is her diagnosis of bad health tech. Non-clinical founders around her, she says, routinely arrive with "this technology, what's the use case or problem we can fit it into?", which she calls "the most common dilemma that technologists have" and the origin of poor products: "If you have solution first, plug it into a problem. That's how you get kind of EMRs where essentially they're billing systems. We've tried to plug them in as clinical solutions and that's why they don't really match physicians workflows at all" [1]. She also insists the bar for a worthwhile problem is low and personal: "If you have a moment where you have a task and you're like, this is hurting me on some level, that's it. That's the that you need to fix" [1].

On medical students as the best problem identifiers

She makes an unusual case for trainees specifically. Medical students "might be some of the most well-positioned people to start healthcare technologies because not only are they not sort jaded by having used things for a long time, but they uh have that nascent mindset" and can identify problem spaces themselves [1]. Her own clipboard ideas ranged from giving residents iPads so that the notes taken while running the list are captured rather than thrown away, which she describes as "extremely valuable data that we were just tossing out every day", to adding post-visit touch points where patients are called to review their medications, which became part of Simbie AI's product [1]. She reports that medical students now reach out to her about alternate paths in medicine, a topic she says she feels very passionately about [1].

On what clinical specificity buys you in a product

Khosla positions her contribution to her own team as domain precision rather than technology. She went into entrepreneurship saying "I'm a physician and I have idea, I know the problems. I know what I saw is breaking and I know how it's hurting my physician colleagues and I definitely see how it's hurting patients" [1]. Technologists, she allows, can get as far as a general ambition such as improving patient intake, but what she brought was the specific: medication reconciliation, which in clinic she saw consume "sometimes 40 % of the visit", down to the level of "What meds are you taking? Which tubes? Which creams? How many times a day? Which one did you stop? Why?", all of it gatherable in advance to improve the in-clinic experience [1]. Scheduling is another process she treats as far more complex than it looks and consequential for patient care [3].

Simbie AI itself she describes as generative AI voice agents that use large language models for human-like speech and voice synthesisers for human-like tone, automating tasks "as if they're your medical staff in a medical practice" [1]. The concrete surface: answering all phone calls 24/7 in multiple languages at the front desk, and calling patients to report biopsy results and review patient education [1]. She also discusses AI's potential to transform dermatology and improve patient outcomes [3].

On incubators and how physicians should learn to build

She is unreserved about the value of accelerators for clinicians without business or technology backgrounds. Y Combinator's contribution, in her telling, was everything downstream of the insight: "Okay, you know the problem. Now let us give you the ropes on how you turn this, how you interact with the startup community, first of all, how you meet the right people and how you turn your idea into MVP, minimum viable product" [1]. After that, she says, the entire journey is "execution, iteration, testing, iteration, testing, iteration" [1]. She names TechStars as another well-known option and points to a growing set of programmes affiliated with health systems [1][3].

Her preferred future is more structural. In her "ideal world, there would be incubators actually in health systems and in uh academic medical centers" [1]. The reasoning comes from rounds: medical students, physicians, PAs, NPs, pharmacists and nurses already work in interdisciplinary teams and all of them identify problems daily. She wants a recurring interdisciplinary forum, monthly, where people bring ideas forward, with engineers present to help break a problem "down into an engineering level execution nugget" and ideally investors in the room too, which she believes would generate the most relevant solutions right inside the academic medical centre [1][3]. Making these resources visible and immediately accessible to trainees and practising physicians is something she frames as a goal to build with other clinician-innovators, a community she describes as small but wants to grow [1].

Takeaways

• The problem space outlasts the product: expect the solution to change constantly, since "There's literally a new voice AI model coming out essentially every four days right now" [1].
• Fewer than 5% of major health system CEOs are physicians, a decline she sees mirroring the corporatisation of the field [1].
• Cost-saving and throughput gains hide an uncounted cost: burnout in a fifth of a department and early retirements that never appear on a P&L [1].
• Solution-first thinking produces tools like EMRs that are really billing systems and therefore do not match physician workflows [1].
• Medical students are strong candidates to found health tech because they are unjaded and can name point-of-care problems firsthand [1].
• Medication reconciliation can consume roughly 40% of a clinic visit and can largely be gathered before the patient arrives [1].
• Simbie AI's voice agents handle 24/7 multilingual front-desk calls and outbound calls for biopsy results and patient education [1].
• Put incubators inside academic medical centres, with engineers and investors joining interdisciplinary teams that already spot problems every day [1][3].