“The AI Is Finally Good Enough”: Why Now Is the Moment for AI in Healthcare
Resources
December 4, 2025
If the last few years in healthcare tech have felt like endless hype cycles, with each new AI product promising to transform clinical work, it’s clear something different is happening right now. For the first time, the technology has caught up to the ambition.
Our marketing team sat down with Joseph Tutera, our founder and CEO, to talk about why now is the moment when AI in healthcare is delivering real, tangible value.
CarePilot: Everyone seems to be talking about AI in healthcare, and quite frankly we’re getting the feeling that people are sick of it. However, it feels like there’s something different this time. What changed?
Joseph: What’s changed is simple but massive: the underlying models are finally good enough.
If you zoom out, the real breakthroughs didn’t come from healthcare-specific innovation. They came from the core AI research community. This is companies including but not limited to OpenAI, Anthropic, Google DeepMind and others. Companies that have spent the last 10+ years grinding.
For the first time, we have systems that can generalize across an enormous range of tasks instead of just pattern matching. And that means in healthcare, you can now rely on an AI system to handle real administrative workloads with a level of accuracy that simply wasn’t possible even two years ago.
The “dirty little secret,” if you want to call it that, is that the AI is just good enough now. Healthcare didn’t make it good enough: the general models did. The question now is which companies can actually apply them in a way that makes clinicians’ lives better.
CarePilot: So this isn’t about “healthcare AI” being a unique field, right? It’s about timing and application?
Joseph: Exactly. I don’t buy the idea that there’s a “healthcare foundation model.” That’s marketing spin. Unless you’re doing frontier work like Google’s AlphaFold and looking into true scientific modeling or diagnostics, you’re not building a better core model than the frontier labs.
Most of healthcare is knowledge work. Writing notes, generating codes, staging orders. All of that is structured knowledge work, and the frontier models are already phenomenal at it.
So the smart move isn’t to waste years trying to pretrain your own “healthcare LLM.” The smart move is to accept that the models are powerful and focus on everything around them: workflow design, user experience, integration, and support. That’s where durable value lives.
CarePilot: You’ve said that it’s actually an advantage to be a company founded in 2023. Why?
Joseph: People sometimes say to us, “Oh, you started in 2023. So you’re the new kids.” Look, I get it. In most industries, you’d rather your bank or insurance company be a century old. But in AI? Starting in 2023 is a gift.
You’re building on a completely different foundation. The technology available in 2022 versus 2023 was night and day. Companies that started earlier often have legacy stacks that just can’t take advantage of the new capabilities.
It’s like launching a website before broadband: you built for dial-up. Once broadband arrived, the entire design paradigm changed. Ambient AI in 2023 is broadband.
CarePilot: What about the market itself? Are clinicians, EHRs, and health systems finally ready for this kind of tech?
Joseph: I think we’re right at the inflection point. Within 36 months, every healthcare provider in the U.S. will be using some form of ambient scribe or charting assistance.
EHR vendors are already moving this way. They’re opening SDKs, exploring new integration models, realizing that the future isn’t about living in the EHR, but working around and through it.
Providers can feel it too. They’ve been burned by bad tech before, with EHRs that promised to “save time” and just added clicks. But they can tell this is different. The quality bar has finally been raised.
CarePilot: Ambient AI started as a note-taking tool. What’s next?
Joseph: Note-taking was the first and quite frankly easiest step. The exciting work now is what happens after the note. Think about what happens when the AI codes your visit, manages your problem list, stages orders, even generates referrals. That’s when it shifts from assistant to collaborator.
But execution matters. It’s one thing to say, “Our AI will do all of that.” It’s another to design the workflow so that it actually works, with clean inputs, clear outputs, and intuitive UI. That’s the new frontier: turning capability into experience. The best product doesn’t matter if the user experience is terrible.
Startups that obsess over workflow and user empathy will outpace the incumbents. Independent practices especially care about quality because they live or die by efficiency.
CarePilot: What’s the biggest misconception about ambient AI right now?
Joseph: It’s the claim that “no change is required to adopt x tool.” That’s total nonsense. Any meaningful technology requires behavior change.
When a provider starts using CarePilot, they’re not suddenly free of their EHR. They’re just spending their time differently. Now they’re reviewing the AI’s work instead of typing every note. It’s a better experience, but it’s still a workflow transition.
We treat that transition as part of the product. Onboarding, change management, real human support and the whiteglove experience that surrounds everything we do is what drives adoption. Most vendors skip that or treat it as a cost center.
CarePilot: If you’re a healthcare leader, what’s your takeaway?
Joseph: This is the moment to lean in.
If you were evaluating ambient AI two years ago, it was probably too early. But now, the models are ready, and the best companies are building the right layers around them. Design, integration, and support.
The worst thing you can do is pick the “cheap” option just because it’s cheap. The right move is to choose the partner who’s building for the future, the one doing more than just transcribing notes.
By 2027, the practices that made that bet early will be thriving.