The latest articles on DEV Community by Blake M Lucey (@blakelucey). https://dev.to/blakelucey https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F533917%2F0d3784de-6e47-45e6-8326-b1559c38270d.jpeg https://dev.to/blakelucey en Blake M Lucey Wed, 10 Jun 2026 19:12:07 +0000 https://dev.to/blakelucey/building-healthcare-ai-taught-me-that-the-model-is-the-easy-part-8ef https://dev.to/blakelucey/building-healthcare-ai-taught-me-that-the-model-is-the-easy-part-8ef <p>When most developers think about healthcare AI, they think about models.</p> <p>Which model should we use?</p> <p>Should we use a managed API or self-host?</p> <p>How large does the context window need to be?</p> <p>What's the cost per token?</p> <p>After spending the last several months building healthcare AI software, I've come to a different conclusion:</p> <p>The model is often the easiest part of the system.</p> <h2> The Real Challenge </h2> <p>Building a proof of concept with an LLM has never been easier.</p> <p>You can sign up for an API, send a prompt, and have something impressive running in an afternoon.</p> <p>The hard part begins when you try to turn that proof of concept into something people can actually use.</p> <p>Questions start appearing immediately:</p> <ul> <li>How does this fit into an existing workflow?</li> <li>What happens when the model is wrong?</li> <li>How should users understand the limitations of the system?</li> <li>How do you authenticate requests?</li> <li>How do you monitor usage?</li> <li>How do you manage costs?</li> <li>How do you maintain trust?</li> </ul> <p>In healthcare, these questions become even more important.</p> <h2> Information vs Decisions </h2> <p>One lesson I've learned is that users rarely need more information.</p> <p>They need help making decisions.</p> <p>A patient may already know:</p> <ul> <li>Their symptoms</li> <li>Their insurance provider</li> <li>Their local healthcare options</li> </ul> <p>What they often don't know is:</p> <ul> <li>Whether telehealth is appropriate</li> <li>Whether urgent care is a better option</li> <li>Whether they should wait for a primary care appointment</li> <li>What the likely cost implications are</li> </ul> <p>The challenge isn't generating more text.</p> <p>The challenge is helping people navigate uncertainty.</p> <h2> Building for Reality </h2> <p>Many AI demos assume a perfect environment.</p> <p>Real users don't operate in perfect environments.</p> <p>They are busy.</p> <p>They are distracted.</p> <p>They may not understand healthcare terminology.</p> <p>They may not trust AI recommendations.</p> <p>That means AI systems have to do more than generate answers.</p> <p>They need to provide context, communicate uncertainty, and fit naturally into existing decision-making processes.</p> <h2> What I'm Watching </h2> <p>I think the next wave of healthcare AI products will be less focused on model capabilities and more focused on operationalizing AI within real-world workflows.</p> <p>The winners won't necessarily be the teams with access to the largest models.</p> <p>They'll be the teams that understand users, workflows, and trust.</p> <p>The technology is important.</p> <p>The experience around the technology is what ultimately creates value.</p> <p>What lessons have you learned while moving AI projects from proof of concept to production?</p> ai infrastructure