Polina, co-founder of Phenomenon Studio and product strategist. Building and scaling digital products with $500М+ raised collectively.
getty
In 2025, American digital health startups raised $14.2 billion in venture funding—35% more than the year before. Fifty-four percent of all funding went to AI-enabled digital health companies. The market rewards technological ambition, but a product's viability doesn't depend on ambition alone. Around 90% of healthtech startups fail, and 60% don't survive their first five years.
My team and I do a lot of work designing and redesigning healthech products, and I can say with certainty: Users don't evaluate how an AI model is built or how complex the system is under the hood. What matters to them is how the product fits into their life or professional routine.
In this column, I'll break down when technology starts working against the idea and how to stress-test a solution before development begins.
To see how a tech-first approach can destroy even a well-funded healthtech product, let's go back almost a decade. In 2017, the healthtech startup Forward Health raised $657 million and built CarePod, medical pods providing autonomous diagnostics designed to be "the world's first AI doctor's offices." Inside were body scanners, biometrics, blood testing and AI screening.
But patients weren't rushing to trust their health to a pod in a shopping mall, and the healthcare system wasn't ready for a model with no physician at the center. Forward's revenue never reached $100 million, and in November 2024, it abruptly shut down.
Although Forward was a physical product, its story illustrates a risk that works the same way in digital environments: When a solution is built around what technology can do rather than what people actually need, that technology quickly turns from a tool into an end in itself. Jan Lim, Assistant Chief Operating Officer at Singapore's Ng Teng Fong General Hospital, captured it: "As more 'tech sexy' solutions appear on the market, the temptation grows to look for a problem that justifies the technology—not the other way around."
Picture a doctor's shift: talking to a patient, entering data into an electronic health record, updating prescriptions, all repeated dozens of times a day. According to a study by Tebra, for every 15 minutes of patient interaction, a physician spends an average of nine additional minutes on documentation, and a single documentation task can require 346 clicks.
If you introduce a product into this workflow that isn't integrated into the real process, the doctor has to figure out how the tool works and whether its data can be trusted.
Clinical alerts are a telling example. The idea behind them is sound: warn the clinician about a risk or a dangerous situation in time. But in practice, physicians work with multiple alert-generating systems simultaneously, and the volume of signals quickly becomes overwhelming. There have been cases where physiological monitors generated over two million alerts per month—187 per patient per day.
For patients, complexity manifests as a gap between information and the next action. If a patient portal makes it hard to find the right data and a recommendation isn't tied to a concrete next step, the user experience looks like chaos.
A telehealth platform may promise a symptom checker, automated pre-consultation data collection and smart routing to the right specialist. But if the patient doesn't understand why they were directed to this particular doctor, the technological scenario doesn't relieve anxiety and makes the decision about the visit harder.
The same happens with AI-powered healthtech products. They can explain lab results, prioritize risks or suggest next steps, but for a user to trust the product, they need to both see the recommendation and understand what it's based on.
In our work redesigning healthtech products, we often see that the problem isn't weak technology. Quite the opposite. Products are getting technologically stronger faster than users can figure out why they should trust them. More often, the issue is that they don't meet the user's needs.
That's why before development—or before adding new technology to a product—I'd recommend three stages of validation:
The question isn't "How do we add AI to the product?" but "What specific action is it supposed to simplify?"
For example, if a patient isn't completing a booking, maybe the problem isn't that you need AI triage. Maybe the booking flow is too long.
It's essential to understand the context in which users consume information and make decisions: where the physician looks for data, what gets duplicated manually, where the patient drops off, where support tickets originate, where people leave the product to verify information elsewhere.
To find this out, you can use shadowing, interviews, session analysis, support tickets or current flow mapping. The goal is to find the point where technology removes an unnecessary action or doubt.
Build a simple clickable mockup around a specific task and measure time-to-action. For a physician, this might be the gap between reviewing data and making a clinical decision. For a patient, between understanding their condition and booking a consultation.
In one of our projects, we reduced time-to-action through a next-action logic. Instead of a dashboard where the user hunts for the right section among dozens of statuses, the screen shows one next step. This eliminated unnecessary navigation between sections and shortened the path to action from multiple screens to one.
The key question for every UX decision in healthtech is simple: Does it bring the person closer to getting help, or push them further away?
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。