Why Healthcare AI Keeps Failing — It's Not the AI, It's the Integration

What's really happening with AI in healthcare? The common story is that health systems just need to find the right tool — the best ambient scribe, the smartest chatbot. But the reality is more complicated.

In this episode, I break down why Sutter Health's AI agent deployment through Hyro tells us everything about where healthcare AI is actually heading, why 63% of healthcare leaders say interoperability is the number one AI capability they want, and why the organizations seeing real ROI did the boring infrastructure work first.

If you're a benefits consultant, health system leader, or anyone advising healthcare organizations — stop evaluating AI tools. Start evaluating integration readiness.

Sources and the deep dive: danmccoymd.substack.com

Transcript

SPEAKER_00 0:07

So, what's really happening with AI and healthcare? The common story is that health systems just need to find the right AI tool, the best ambient scribe, the smartest chatbot, the most accurate diagnostic model. But the reality, it's way more complicated. I met with a benefits vendor last month who's still pitching single solution AI improvements. Think better, prior off, smarter scheduling, point solutions for point problems. And I realized that's the wrong strategy entirely. Here's what I'm going to share in this video. First, why Sutter Health's new AI agent deployment tells us everything about where this is actually heading. Second, why AI without interoperability is just noise and what that means for anyone buying or building healthcare AI right now. And third, the infrastructure mindset that separates health systems that are actually getting results from those that are just burning money on pilot projects. For benefit consultants, health system leaders, or anyone advising healthcare organizations, this is the conversation you need to be having in virtually every meeting. The common story with healthcare AI is that you find a problem, say call center wait times, and you just bolt on an AI solution. But here's what's actually happening at Sutter Health. They just deployed AI agents through a company called Hero or HIRO, not sure how to pronounce it. These agents handle appointment scheduling, prescription refills, billing inquiries, the full spectrum of a routine patient interaction. Here's the part that most people miss. This isn't a chatbot project. Jennifer Bollinger, Sutter's chief consumer and brand officer, said the experience should feel easy and effortless, regardless of whether patients are seeking information via chat, voice, or text. That word, regardless, is doing a lot of work here. They're not deploying separate AI for chat, separate AI for voice, separate AI for text. They're deploying one cohesive experience across all the channels. And they're measuring success not by how well AI performs in isolation, but by customer satisfaction scores, first call resolution rates, and whether staff can focus on complex cases instead of routine ones. The AI Governance Council at Sutter is overseeing this with strict data sourcing, explainability requirements, and compliance guardrails. So the key insight is the AI is the tip of the iceberg. The real investment is the integration layer, the governance framework, and the experience design that makes it all seamless. And here's where it really gets interesting. A recent survey found that 63% of healthcare leaders say interoperability is the single most appealing AI functionality. Not the smartest algorithm, not the best language processing, but integration readiness. The common story is that AI will transform healthcare by being really, really good at specific tasks. But the reality is that most AI implementations fail not because the AI doesn't work. They fail because the AI can't access the data it needs to work. One healthcare CIO put it this way, and I'm paraphrasing here, AI without interoperability is just noise. Think about what an AI agent needs to actually help a patient with a billing question, access to the patient's account, visibility into their insurance information, integration with EHR for clinical context, connection to the billing system for payment options, real-time data exchange to avoid stale information that's outdated. If any of those connections are broken, the AI either hallucinates an answer or punts to a human, and you've gained nothing. The organizations getting real results are the ones that are building what I call integration-first architecture. API-based connections using FHIR standards, aggregated data repositories, real-time data exchange between ERP, EMR, and third-party platforms. And critically, governance frameworks that can handle AI's broader access to patient data. That's crucial. So the takeaway is this before you evaluate any AI vendor, you need to evaluate your own integration readiness. If the data foundation isn't there, the AI really won't matter. This brings me to the benefits vendor meeting I mentioned. They were pitching an AI solution for prior authorization. Now on paper, it looks great. Faster approvals, less admin burden. But when I asked how it integrates with the client's existing systems, the answer was basically we'll figure that out during implementation. That's a red flag. The organizations that are actually seeing ROI from Healthcare AI 2026 are the ones that did the boring work first. Clean, high-quality data, robust governance frameworks, process validation before automation, think cloud-first architecture. As one expert put it, data is the lifeblood not only of operations, but of enhanced AI. I'm going to be honest with you. Most healthcare organizations aren't there yet. And that's okay. But it means the conversation shouldn't be which AI tool should I buy? It should be what foundations do we need to build before AI can actually help us. The part most people miss is that AI implementation isn't a project, it's a capability. It requires ongoing data hygiene, system connectivity, and governance evolution. If you're a benefits consultant, this changes how you evaluate clients. Stop evaluating point solutions. Start evaluating integration readiness. If you're a health system leader, this changes your investment priorities. The unsexy infrastructure work, so think APIs, data governance, interoperability. That's where the real competitive advantage is being built. So let me pull all of this together. We covered three things. First, Sutter Health shows that successful AI deployment is about unified experiences across channels, not standalone tools. The AI is just the surface. The integration and governance underneath is what makes it work. Second, AI without interoperability, it's just noise. 63% of healthcare leaders agree with me. Integration readiness is the most important capability you should be focusing on. And third, foundation beats hype. The organizations winning with AI did the boring data and infrastructure work first. For anyone advising healthcare organizations right now, the opportunity is huge. Most vendors are still selling point solutions to organizations that need platform thinking. The risk is continuing to chase individual AI tools while competitors are building the integration layer that's actually going to make this all work. If this was helpful, subscribe for more Healthcare AI strategy. And if you want the deeper research on this, including the interoperability frameworks and Sutter Health Case Study, I've got a full write up on my Substack links in the description.