Podcast Summary: Just Now Possible

Episode: When Trust Is Everything: Building AI for Physicians at Healio

Host: Teresa Torres
Guests:

Jen, SVP Product Development
Casey Utley, Senior UX Designer
Matt, VP of Technology
Date: January 22, 2026

Episode Overview

This episode dives deep into the process behind the creation and evolution of Helio AI, the first AI product by Healio, a company serving healthcare professionals with news, education, and clinical guidance. The team discusses identifying pain points for physicians (especially information overload and time scarcity), experimenting with AI-powered solutions, and the paramount importance of trust and transparency in the medical space. The conversation covers early prototyping, technical architectures, building trust, integration of vetted data sources, human and AI-driven evaluations, and continuous feedback loops.

Key Discussion Points & Insights

1. Problem Space: Information Overload and Time Pressure

Helio AI aims to be a "one-stop shop" for physicians inundated with ever-changing guidelines and studies, enabling fast, reliable access to up-to-date information.
The crisis point: Physicians' limited time, especially at the point of care, makes rapid, trusted answers essential (04:00).

Quote:

“If you think about when a provider is at the point of care... they have a question, they need the answer quickly, but they need to make sure it’s coming from a trustable, credible source.”
– Jen (04:10)

2. Trust: The Overarching Theme

Physicians need not only accurate answers, but transparency about sources—especially in a landscape with thousands of journals varying in credibility (06:45).
Helio surveyed 300+ healthcare professionals (05:54) and conducted continuous usability sessions to understand attitudes toward AI and pain points around trust.

Quote:

“We know that they really value the trusted sources, trusted medical sources, and the transparency of tools like this, and they’re willing to use [AI] at point of care when it is transparent for them.”
– Casey (06:02)

3. Discovery, Prototyping, and Usability Insights

Initial prototypes let physicians use freeform prompts. Surprisingly, many asked about patient communication and empathy, not just diagnosis/treatment (07:54).
This shifted product tone—supporting not just clinical accuracy, but also emotionally sensitive, empathetic explanations.

Quote:

“They started to enter in prompts about patient communication... they were asking ‘How do I explain this diagnosis to my patient?’ That actually shifted a direction, too, with our product...”
– Casey (08:00)

The tool is mostly used in preparation for appointments, not during live patient encounters—typically on breaks or the day before, to quickly prepare for important conversations (09:14).

4. Architecture & Technology Choices

Product Flow:

Physician enters a prompt → System pulls relevant data via RAG & search → AI synthesizes a concise, source-annotated answer → Citations allow for deep dives; interface supports both summaries and details.

Key Technical Choices:

Heavy focus on RAG (Retrieval-Augmented Generation) with careful curation of input sources:
- Internal: Helio news, CME, guidelines
- External: PubMed (most trusted journals), others by strict vetting (13:00, 26:40, 27:23)
Multiple retrieval strategies: Lexical, vector, and semantic searches, tuned for different data types and to surface the most recent and relevant content (23:33, 30:09).
Continuous performance tuning for both speed and trustworthiness, balancing trade-offs—e.g., showing contextual ads while AI processes queries to enhance perceived responsiveness (22:06–23:02).

Quote:

“We do use a combination of [search methods]... There’s definitely a lot of gotchas than just saying, ‘Okay, I have a vector database and it’s just going to do its thing and give me the best answers.’ There’s a lot of tuning and a lot of examining what users are actually asking and then refining. And it’s a continuous loop that I don’t think will ever end.”
– Matt (23:33)

5. Building Trust with Transparency & Interface Design

Early and continuous feedback cycles with clinicians were critical.
UI supports progressive disclosure—a summary with subscripts for every statement, enabling source verification (31:33).
On-screen references allow physicians to see and verify information levels as needed (32:44).
The product is HIPAA-compliant and avoids storing any personally identifying information (19:32, 36:15).

Quote:

“Trust... is built through transparency, tone adjustment, and just the respect for our users' time.”
– Casey (46:12)

6. Human and AI Feedback Loops

Real-world feedback from physicians is integral:
- Integrated thumbs up/down with reasons
- Deep involvement of advisory boards and a “Helio Innovation Partners” group (34:45–35:43)
Early feedback focused on speed, tone, and formatting improvements: e.g., adding more bullet points and patient-facing empathy.

Quote:

“We have a team that meets weekly, and we are just continuing to work on improvements and prioritize what’s important... if there’s changes that we need to make on the back end or tweaks... we’re prioritizing those.”
– Jen (38:26)

AI as “judges”: Experimenting with LLM-based evals for safety, accuracy, relevancy, etc., but always validating against human evaluation (39:36). Still, feedback from real users is paramount.

7. Guardrails, Compliance & Evaluation

Upfront guardrails filter inappropriate or unsafe queries and redact any PHI before processing, critical for HIPAA compliance (36:15).
Multiple evaluation strategies are being tested, including “LLMs as judges” for attributes like safety, accuracy, and faithfulness, but the team does not rely solely on AI feedback—real clinician feedback is decisive (39:36–41:46).

Quote:

“Trust, again, is the most Important thing. And we don’t want to just blind trust the mail.”
– Matt (37:41)

8. Evolution, Experimentation, and Lessons Learned

The architecture and evaluation strategies are evolving; there’s a strong ethos of experimentation with both technical approaches and feedback systems.
There is “no one right answer”—success is context-dependent and hinges on tight feedback with the end-user (43:38–44:36).

Quote:

“You really have to experiment to figure out what works in your context... there’s so many ways to do things and the way that’s going to work is going to be very context dependent.”
– Teresa (44:23)

9. Key Future Directions

Double down on discovery, maintaining close collaboration with physicians (48:00).
Scaling continuous improvement through feedback and advisory boards.
Remaining focused on “responsible adoption of AI” and protecting the trust between clinicians and their patients (45:32).

Quote:

“AI designing for AI, it doesn’t reduce the need for discovery at all... The more responsibility a product has, especially in the healthcare space, the more deeply we have to listen to our users... Trust... is built through transparency, tone adjustment, and just the respect for our users' time.”
– Casey (46:04–46:43)

Memorable Moments & Notable Quotes

On empathy and product shift:

“Physicians... were actually looking for a more empathetic response from our product.”
– Casey (08:00)
On building rapidly with new tools:

“With these tools you can make the product…in a weekend.”
– Matt (16:34)
On the need for experimental architecture:

“Maybe a year ago, all of the KPIs were based upon speed…trustworthy actually also brings with it a little bit less speed because you have to make sure that the answer is accurate and you actually are going to the best sources.”
– Matt (25:04)
On continuous evolution:

“It’s so true in the times of AI that It’s really easy to build anything, but the design aspect... has never been more important.”
– Matt (47:20)

Important Timestamps & Segments

[02:14] – Problem statement: information overload for physicians
[05:54] – Importance of trust and transparency; clinician survey
[07:54] – Uncovering unexpected needs: communication, empathy
[09:14] – How and when physicians use Helio AI
[12:58–14:04] – Curating sources and building trust, why not just use ChatGPT
[16:34] – Building the prototype in a single weekend
[22:06–23:02] – Product architecture overview, use of contextual ads in waiting periods
[31:33–34:19] – UI for trust: interface, citations, and progressive disclosure
[36:15–37:41] – Guardrails, masking PHI, and human vs. AI evals
[45:32–46:43] – Commitment to responsible AI, discovery, and partnership with users
[48:00] – Looking ahead: continuous validation and roadmap

Conclusion

This episode offers an inside look at how a legacy healthcare company is navigating rapid change through responsible, iterative AI adoption. The journey of Helio AI is rooted in constant engagement with physician end-users, deliberate technology choices prioritizing trust and transparency, and a culture of experimentation with feedback loops at every stage. For anyone building AI for high-trust domains, the Helio AI story provides vital lessons in user-centricity, prototyping, and the future of AI-powered tools in clinical workflows.

Podcast Summary: Just Now Possible

Episode: When Trust Is Everything: Building AI for Physicians at Healio

Host: Teresa Torres
Guests:

Jen, SVP Product Development
Casey Utley, Senior UX Designer
Matt, VP of Technology
Date: January 22, 2026

Episode Overview

Key Discussion Points & Insights

1. Problem Space: Information Overload and Time Pressure

Helio AI aims to be a "one-stop shop" for physicians inundated with ever-changing guidelines and studies, enabling fast, reliable access to up-to-date information.
The crisis point: Physicians' limited time, especially at the point of care, makes rapid, trusted answers essential (04:00).

Quote:

“If you think about when a provider is at the point of care... they have a question, they need the answer quickly, but they need to make sure it’s coming from a trustable, credible source.”
– Jen (04:10)

2. Trust: The Overarching Theme

Physicians need not only accurate answers, but transparency about sources—especially in a landscape with thousands of journals varying in credibility (06:45).
Helio surveyed 300+ healthcare professionals (05:54) and conducted continuous usability sessions to understand attitudes toward AI and pain points around trust.

Quote:

“We know that they really value the trusted sources, trusted medical sources, and the transparency of tools like this, and they’re willing to use [AI] at point of care when it is transparent for them.”
– Casey (06:02)

3. Discovery, Prototyping, and Usability Insights

Initial prototypes let physicians use freeform prompts. Surprisingly, many asked about patient communication and empathy, not just diagnosis/treatment (07:54).
This shifted product tone—supporting not just clinical accuracy, but also emotionally sensitive, empathetic explanations.

Quote:

“They started to enter in prompts about patient communication... they were asking ‘How do I explain this diagnosis to my patient?’ That actually shifted a direction, too, with our product...”
– Casey (08:00)

The tool is mostly used in preparation for appointments, not during live patient encounters—typically on breaks or the day before, to quickly prepare for important conversations (09:14).

4. Architecture & Technology Choices

Product Flow:

Physician enters a prompt → System pulls relevant data via RAG & search → AI synthesizes a concise, source-annotated answer → Citations allow for deep dives; interface supports both summaries and details.

Key Technical Choices:

Heavy focus on RAG (Retrieval-Augmented Generation) with careful curation of input sources:
- Internal: Helio news, CME, guidelines
- External: PubMed (most trusted journals), others by strict vetting (13:00, 26:40, 27:23)
Multiple retrieval strategies: Lexical, vector, and semantic searches, tuned for different data types and to surface the most recent and relevant content (23:33, 30:09).
Continuous performance tuning for both speed and trustworthiness, balancing trade-offs—e.g., showing contextual ads while AI processes queries to enhance perceived responsiveness (22:06–23:02).

Quote:

“We do use a combination of [search methods]... There’s definitely a lot of gotchas than just saying, ‘Okay, I have a vector database and it’s just going to do its thing and give me the best answers.’ There’s a lot of tuning and a lot of examining what users are actually asking and then refining. And it’s a continuous loop that I don’t think will ever end.”
– Matt (23:33)

5. Building Trust with Transparency & Interface Design

Early and continuous feedback cycles with clinicians were critical.
UI supports progressive disclosure—a summary with subscripts for every statement, enabling source verification (31:33).
On-screen references allow physicians to see and verify information levels as needed (32:44).
The product is HIPAA-compliant and avoids storing any personally identifying information (19:32, 36:15).

Quote:

“Trust... is built through transparency, tone adjustment, and just the respect for our users' time.”
– Casey (46:12)

6. Human and AI Feedback Loops

Real-world feedback from physicians is integral:
- Integrated thumbs up/down with reasons
- Deep involvement of advisory boards and a “Helio Innovation Partners” group (34:45–35:43)
Early feedback focused on speed, tone, and formatting improvements: e.g., adding more bullet points and patient-facing empathy.

Quote:

“We have a team that meets weekly, and we are just continuing to work on improvements and prioritize what’s important... if there’s changes that we need to make on the back end or tweaks... we’re prioritizing those.”
– Jen (38:26)

AI as “judges”: Experimenting with LLM-based evals for safety, accuracy, relevancy, etc., but always validating against human evaluation (39:36). Still, feedback from real users is paramount.

7. Guardrails, Compliance & Evaluation

Upfront guardrails filter inappropriate or unsafe queries and redact any PHI before processing, critical for HIPAA compliance (36:15).
Multiple evaluation strategies are being tested, including “LLMs as judges” for attributes like safety, accuracy, and faithfulness, but the team does not rely solely on AI feedback—real clinician feedback is decisive (39:36–41:46).

Quote:

“Trust, again, is the most Important thing. And we don’t want to just blind trust the mail.”
– Matt (37:41)

8. Evolution, Experimentation, and Lessons Learned

The architecture and evaluation strategies are evolving; there’s a strong ethos of experimentation with both technical approaches and feedback systems.
There is “no one right answer”—success is context-dependent and hinges on tight feedback with the end-user (43:38–44:36).

Quote:

“You really have to experiment to figure out what works in your context... there’s so many ways to do things and the way that’s going to work is going to be very context dependent.”
– Teresa (44:23)

9. Key Future Directions

Double down on discovery, maintaining close collaboration with physicians (48:00).
Scaling continuous improvement through feedback and advisory boards.
Remaining focused on “responsible adoption of AI” and protecting the trust between clinicians and their patients (45:32).

Quote:

“AI designing for AI, it doesn’t reduce the need for discovery at all... The more responsibility a product has, especially in the healthcare space, the more deeply we have to listen to our users... Trust... is built through transparency, tone adjustment, and just the respect for our users' time.”
– Casey (46:04–46:43)

Memorable Moments & Notable Quotes

On empathy and product shift:

“Physicians... were actually looking for a more empathetic response from our product.”
– Casey (08:00)
On building rapidly with new tools:

“With these tools you can make the product…in a weekend.”
– Matt (16:34)
On the need for experimental architecture:

“Maybe a year ago, all of the KPIs were based upon speed…trustworthy actually also brings with it a little bit less speed because you have to make sure that the answer is accurate and you actually are going to the best sources.”
– Matt (25:04)
On continuous evolution:

“It’s so true in the times of AI that It’s really easy to build anything, but the design aspect... has never been more important.”
– Matt (47:20)

Important Timestamps & Segments

[02:14] – Problem statement: information overload for physicians
[05:54] – Importance of trust and transparency; clinician survey
[07:54] – Uncovering unexpected needs: communication, empathy
[09:14] – How and when physicians use Helio AI
[12:58–14:04] – Curating sources and building trust, why not just use ChatGPT
[16:34] – Building the prototype in a single weekend
[22:06–23:02] – Product architecture overview, use of contextual ads in waiting periods
[31:33–34:19] – UI for trust: interface, citations, and progressive disclosure
[36:15–37:41] – Guardrails, masking PHI, and human vs. AI evals
[45:32–46:43] – Commitment to responsible AI, discovery, and partnership with users
[48:00] – Looking ahead: continuous validation and roadmap