AI-Powered Clinical Decision Support Platform for Healthcare Providers

healio · Healthcare · 2026 · source Agentic: no · promotional bias: med · source-backed · extracted with prompt 0.5.4 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: Clinical decision support and medical information retrieval

Healio wanted to help physicians answer point-of-care questions quickly without relying on general-purpose AI systems that hallucinate or lack trusted medical sourcing.

Constraints:

Architecture

Healio AI uses a medical RAG architecture: physician questions retrieve proprietary clinical content and trusted PubMed sources, journal-quality filtering shapes context, an LLM synthesizes cited answers, and physician feedback continuously tunes the product.

Diagram confidence: reconstructed (reconstructed from prose).

flowchart LR
  physician-query-interface["Physician query interface"]
  rag-retrieval-layer["RAG retrieval layer"]
  journal-quality-filter["Journal-quality filter"]
  llm-synthesis-layer["LLM synthesis layer"]
  citations-and-sources["Citations and sources"]
  physician-feedback-loop["Physician feedback loop"]
  innovation-partner-advisors["Innovation partner advisors"]
  physician-query-interface --> rag-retrieval-layer
  rag-retrieval-layer --> journal-quality-filter
  journal-quality-filter --> rag-retrieval-layer
  rag-retrieval-layer --> llm-synthesis-layer
  llm-synthesis-layer --> citations-and-sources
  citations-and-sources --> physician-feedback-loop
  innovation-partner-advisors --> physician-feedback-loop
  physician-feedback-loop --> rag-retrieval-layer
Drag to pan · scroll or use +/− to zoom · reconstructed from the source's described component topology

Components

Component Role Interacts with
Physician query interface Receives point-of-care medical questions from clinicians. RAG retrieval layer
RAG retrieval layer Retrieves Healio content, continuing medical education, and PubMed journal context. Journal-quality filter, LLM synthesis layer
Journal-quality filter Filters lower-quality journals before inference. RAG retrieval layer
LLM synthesis layer Produces concise answers and source-backed explanations for physicians. Citations and sources
Citations and sources Expose transparent medical grounding and source depth. Physician feedback loop
Physician feedback loop Collects thumbs up/down, categories, typed explanations, and advisory-board input. RAG retrieval layer
Innovation partner advisors Provide usability testing and clinical review for product evolution. Physician feedback loop

Data flow

The system ingests Healio clinical content, continuing education, and trusted PubMed journals through RAG, filters lower-quality journals, and returns cited answers to physicians.

“we need to use a rag system in order to not only ingest all of our clinical content from all the news that we have within our site, all the continuing medical education that we have in our site, but we also pull from trusted journals in PubMed.”transcript, stated
“we do have a mechanism that weeds out some of the lower quality journals and pull in only the most respected journals to use during inference to ultimately use for our users.”transcript, stated

Why this architecture

The architecture avoids general-purpose chatbot behavior because physicians need transparent trusted sources and hallucination resistance at point of care.

“Chat GPG when it first launched was laughably bad actually with constant hallucination. So it truly could not be trusted.”transcript, stated
“they really value the trusted sources, trusted medical sources, and the transparency of tools like this, and they're willing to use that at point of care when it is transparent for them.”transcript, stated

Models used: large language model · RAG system

Scale

Alternatives considered (roads not taken)

General-purpose AI answers without medical retrieval and citations

Rejected because: Healio chose RAG, journal filtering, source transparency, and physician feedback because generic chatbots were not trusted for clinical answers.

“Chat GPG when it first launched was laughably bad actually with constant hallucination. So it truly could not be trusted.”transcript, stated
“they really value the trusted sources, trusted medical sources, and the transparency of tools like this, and they're willing to use that at point of care when it is transparent for them.”transcript, stated

Medical answers without source transparency

Rejected because: Physicians valued trusted medical sources and transparency for point-of-care use.

“they really value the trusted sources, trusted medical sources, and the transparency of tools like this, and they're willing to use that at point of care when it is transparent for them.”transcript, stated

Only diagnostic or treatment question support

Rejected because: Prototype testing showed physicians also asked about patient communication, changing tone and product design.

“the physician, they weren't just asking diagnostic or treatment questions. Like I said, they were asking more about patient communication. So that helped us adjust the tone.”transcript, stated

Patterns

rag-retrieval (retrieval)

Healio grounds answers in its own clinical content plus trusted PubMed journals instead of letting the model answer from general memory.

“we need to use a rag system in order to not only ingest all of our clinical content from all the news that we have within our site, all the continuing medical education that we have in our site, but we also pull from trusted journals in PubMed.”transcript, stated

retrieval-ranking (retrieval)

Healio filters journal sources so lower-quality journals do not enter the inference context.

“we do have a mechanism that weeds out some of the lower quality journals and pull in only the most respected journals to use during inference to ultimately use for our users.”transcript, stated

human-in-the-loop-gate (safety)

Physicians remain involved through continuous feedback, answer ratings, and advisory-board input rather than the product being validated only internally.

“we're constantly having those touch points with our users even now that it's launched. Feedback right from the answer, too. So, if someone isn't happy, there's a thumbs up, thumbs down really quickly.”transcript, stated

data-quality-loop (data)

Live physician feedback and selected dissatisfaction categories feed product improvement after launch.

“If someone is unhappy with the answer, there's an option to click on some pre-selected categories of why or they can type in explanation if they choose to.”transcript, stated

Tradeoffs

How to ground point-of-care answers.

Chose curated RAG over trusted clinical sources over general-purpose ChatGPT-style answers.

Rationale: The product needed physician trust, transparent sourcing, and journal-quality control.

“they really value the trusted sources, trusted medical sources, and the transparency of tools like this, and they're willing to use that at point of care when it is transparent for them.”transcript, stated

How to monetize the assistant without blocking the clinician workflow.

Chose contextual ads while the streamed response is loading over making physicians wait on an empty loading state.

Rationale: Healio uses prompt intent to show relevant advertising during response latency.

“We try to associate the ad with keywords that might exist within the prompt. So for instance, if a user is asking about lung cancer, then perhaps there's a lung cancer ad”transcript, stated

Gotchas (what bit them)

General-purpose chatbots were not trusted enough for clinical answers.

Root cause: Early ChatGPT hallucinated and did not expose credible medical sourcing.

Resolution: Use RAG over curated clinical content and trusted journals with transparent citations.

“Chat GPG when it first launched was laughably bad actually with constant hallucination. So it truly could not be trusted.”transcript, stated

The team initially underestimated how physicians would use the tool.

Root cause: Prototype users asked about patient communication, not only diagnosis or treatment.

Resolution: Use moderated usability testing and real prompts to adjust tone, answer format, and source-depth behavior.

“the physician, they weren't just asking diagnostic or treatment questions. Like I said, they were asking more about patient communication. So that helped us adjust the tone.”transcript, stated

Physicians need both short answers and source depth.

Root cause: Point-of-care users are pressed for time but still need to inspect evidence when the question matters.

Resolution: Provide summarized responses while letting users deep dive into cited sources.

“they're pressed on time, but they really do appreciate bulleted summarized responses in our LLM, but they also appreciate going into a deep dive into the sources as well.”transcript, stated

Evaluation

Healio surveyed physicians, ran moderated usability testing, tested a working prototype with real prompts, and continued feedback collection after launch.

Metrics: over 300 healthcare professionals surveyed · about 75% of surveyed respondents reported using AI for patient interactions and patient care

“From the very beginning, we surveyed over 300 healthcare professionals to understand like their attitudes toward AI.”transcript, stated
“In our survey data, we found about 75% of those 300 respondents that Casey had mentioned earlier report using AI um for patient interactions and patient care.”transcript, stated

Outcomes

Podcast transcript with company participants; it gives clear design evidence but limited quantitative post-launch outcome metrics.

“it was a working prototype so that the users can enter real prompts and get real answers and that's where our assumptions really got challenged.”transcript, stated
“As part of that, there's what came from what sources so we can connect up to these citations.”transcript, stated