1 of 17

The Change Agent

A Digital Health Companion for Patients & Families

May 27, 2026

2 of 17

Team

Dr. Rob Quinn

Dr. Gabe Fabreau

Dr. Troy Pederson

Donovan Pederson

3 of 17

Proposed Product

A conversational intake agent that interacts with patients who have been referred to a medical clinic to capture important pre-visit information.

The patient follows a link on their smartphone or device and is taken into a conversational interface where they speak to the agent. Essential, patient-reported information is captured, packaged in a usable format, and passed to the provider in advance of the clinic visit.

WHAT DO WE OFFER

4 of 17

Referral

First Clinic Visit

“Pre-Visit Period”

With the intake agent, we are targeting the time period between referral to a medical clinic and the first clinic visit where they see the healthcare provider.

When a patient is referred to a clinic — particularly to subspecialty or specialist offices — they often face a long wait between referral and appointment, with little or no communication from the office in the interim.

When they finally see the provider, they spend a large part of the visit collecting patient information that could have been captured prior to the visit and made available to that provider

This would allow them to focus on talking about diagnosis, treatment options, and answering questions – in other words, prioritizing the parts of the visit that require the expertise of the provider and are most important to patients.

The silence between referral and visit is wasted time and it could be used to capture data that are essential to the clinical encounter before the visit, keep patients engaged, educate them, etc.

Where intake tools exist today, they are largely inadequate. Paper-based and web-based intake tools tend to be clumsy, poorly designed, difficult to use, and time-consuming.

Almost without exception, they are built for patients who speak English and are comfortable navigating English-language forms.

More recently, people have been leveraging technology to automate the intake process and we will discuss those solutions in more detail when we discuss the competition.

5 of 17

Why are we collecting this information?

What is the best source of that information?

Limited English proficiency

Pre-Visit Period = missed opportunity

Fragmented care & records

High Quality Patient Data

Clinical Care

Quality Improvement

Research

So we have this pre-visit period that is a missed opportunity to engage patients to capture important information in advance of the first visit.

Historically, we have viewed this simply as the transfer of information from patient to provider without much attention paid to the quality of that information, the reason we are collecting it, or what the most reliable source is. In our view, these are very important considerations. High quality patient data, collected from the most reliable source, and collected with purpose is critical to …

Intake agents have shifted the data collection form the clinic visit (clinician-guided and supplemented by chart) to the patient (patient-report) and that may not be a good idea.

High-quality patient information is essential, and not for a single reason.

It is essential for clinical care — clinicians cannot make sound decisions without it.
It is essential for quality improvement, because we cannot improve what we cannot reliably measure.
It is essential for research, where the answer to any meaningful clinical question depends on the integrity of the underlying data.

Accomplishing this is made more challenging because care is fragmented and so are patient records and many patients have limited English proficiency

The focus of our project is to take this intake process to the next level – collect high quality data, from the most reliable source, and collect it with clear purpose.

6 of 17

Recap

SUMMARY

There were several key issues raised and discussed during the last presentation that have shaped our thinking:

Technical feasibility
Funding & building a prototype
Competitive landscape

Technical feasibility: concerns raised by Dr. Shapiro last time regarding our planned 3D avatar and had a follow-up meeting that went into the details of why that approach is not ideal and would not likely be financially and technically feasible currently. This has caused us to re-imagine our intake agent as a voice-only conversational AI tool, rather than a 3D avatar. In many ways, this simplifies the build.

Funding and building a prototype: We have funds to finance the build, but were actively seeking funding opportunities. Since our last meeting we have been successful in PAREX grant competition through the university of Calgary – 20k over 2 years. We also have a line on funding from an industry partner that will likely total 150k-175k over two years. Waiting to finalize this.

-We have sought additional advice from Dr. Shapiro and have interviewed and are meeting with a developer with extensive experience in the health care space this week who we plan to hire for the project.

-We elected to do the build ourselves, rather than contract it out (Scott had suggested AltaML)

Competitive landscape: Concerns were raised that this was becoming a very crowded space with players like Microsoft developing intake tools and other players including FirstHx that Scott brought to our attention. At first, this was somewhat deflating and the source of some anxiety, but on closer examination, it was actually very helpful in fleshing out what differentiates us….. The competition has been focused on applying technology to make the process of collecting information from patients faster and more efficiently….

7 of 17

3D Digital Avatar NOT feasible

Technical issues

Cost considerations

Shift to voice-based conversational intake

Technical Feasibility

8 of 17

Patient Mobile Application

Voice

Capture

Audio Stream

Backend API

(FastAPI)

Audio File Storage

MySQL Database

Administrative Dashboard

Clinician

Whisper Speech-to-Text

Text-to-Speech

RAG Vector Database

Fine-tuned LLM

System Architecture

Key Components:

Patient Mobile App: Captures voice input and displays AI responses

Backend API: Coordinates all services and manages data flow

Whisper STT: Converts speech to text in real-time

Fine-tuned LLM: Processes conversation with RAG context for medical accuracy

RAG Vector DB: Provides medical knowledge context to the LLM

TTS: Converts AI text responses to natural speech

MySQL Database: Stores patient records, sessions, and metadata

File Storage: Stores audio files and conversation recordings

Admin Dashboard: Web interface for clinicians to review patient data

Sydney has proposed a 40-week implementation plan to take from beginning to clinical deployment

See document he sent us after meeting – technical plan

This figure is taken from Data Pipeline section

Cost:

$430-1,030/month (API-based)

$80-150/month (self-hosted approach, excluding one-time costs)

9 of 17

Funding

Parex grant: ~20k over 2 years

Private funding: ~$150-175k over 2 years

Pursuing other opportunities

10 of 17

Competition

11 of 17

Talk about how we now have this automation that means you only need a single driver (no longer need 2 guys on back of truck).

This has made it way faster to collect garbage and less labour intensive with 1 person doing the work of 3.

BUT

You’re still collecting garbage…. And this is analogous to how technology is being applied to the intake process.

They are making the process of data collection faster and more efficient…..but it’s still garbage….they are solving the wrong problem in our view

The trick is converting that garbage into something useful…. We are increasingly drowning in data and yet starving for usable information.

The real problem is the quality of the data and making sure that what we collect is collected with a clear purpose and is usable. Not just checking a box.

12 of 17

A Revised Workflow

Intake Agent

Chart Synthesis

Clinic Visit

Digital Health Passport

Differentiator is not the conversational interface per se, it’s the design philosophy behind it.

Not all of the information that is required for the first clinic visit is equally well captured from the same source.

Some elements are reliably reported by patients themselves, such as basic demographics, social determinants of health are, medication lists and for some elements, the patient is in fact the best — and often the only — authoritative source.

Other elements are not well captured from patients, even when the patient is willing and engaged. Past medical history is the clearest case: patients frequently do not remember what happened to them or when it happened, and often lack the medical vocabulary to convey the details a clinician needs.

Lab results, imaging, and other testing are similarly unreliable when obtained from patient recall. For these elements, the medical record is the appropriate source.

In current practice, prior to the visit, a clinician (or someone on their behalf) reviews the chart and compiles the relevant information into a usable format. That is real, skilled work, and it should not be displaced onto an unreliable patient-facing intake process.

The other intake agents out there have ploughed ahead with capturing everything they can prior to the visit, without any thought given to the reliability or usefulness of that information. This wastes patients’ time and passes information to the provider that isn’t reliable.

This distinction — what patients can reliably self-report versus what should come from the chart — is foundational to the design of the Change Agent.

13 of 17

Multilingual interface

Social determinants of health

Medications & allergies

Intake Agent

Chart Synthesis

Clinic Visit

Digital Health Passport

Step 1 — Intake Agent (pre-visit)

Patients receive a secure link and interact with a conversational interface through their smartphone or device. The goal is not to recreate the entire chart — it is to capture and refresh high-value patient-reported information in a structured and reliable way.

Step 2 — Chart Synthesis (pre-visit)

Separately, the existing medical record is synthesized into a clinically usable format. The standard is not that the information exists somewhere in the chart. The standard is that it is accurate, organized, validated, and useful at the point of care.

Step 3 — The First Visit

Because intake and chart synthesis have already occurred, the clinician’s in-person time can focus on clarification, examination, interpretation, decision-making, and relationship-building. The visit becomes less about data collection and more about clinical care.

Step 4 — Patient-Facing Structured Information

After the visit, the patient receives structured, understandable, validated information derived from the encounter — beginning to build a more portable, patient-accessible longitudinal record that improves continuity across settings

14 of 17

Quality of care inferior

Ability to follow instructions, follow-up, ER visits, medication complications

Impact of trained interpreters

Multilingual interface

Social determinants of health

Medications & allergies

Transportation barriers

Financial barriers

Food insecurity

Errors in medication histories very common (30-70%)

Harm 1.5 M people in US annually

Most errors stemmed from inaccuracies at time of intake

Structured tools help

Flores 2005; Hsueh 2021; Syed 2013; Berkowitz 2016; Tam 2005; Pippins 2008; IOM 2006; Glintborg 2007

There is a very strong clinical and business case for focusing on these three pillars:

Multilingual: Quality of care inferior for patients with limited English proficiency

Trained interpreters impact patient satisfaction and quality of care

Impacts ability to follow instructions, follow-up, ER visits, medication complications

SDOH: Transportation barriers: rescheduled/missed appointments, delayed care

Financial: medication non-adherence

Food insecurity: worse chronic disease outcomes, increased ER / hospitalization use, increased healthcare costs

Medications and allergies: Errors in medication histories very common (30-70%)

Errors harm 1.5M people in US annually

Most medication errors stemmed from inaccuracies at time of intake

If medications are captured using a structured tool, accuracy improves substantially

15 of 17

Current development timeline

2026 Q1

2026 Q2

2026 Q3

2026 Q4

2027 Q1

2027 Q2

2027 Q3

2027 Q4

Concept & Design

Funding Prototype

Milestone

1

Build Prototype

Milestone

2

Milestone

3

Usability Testing & Validation

Clinic Deployment & Testing

16 of 17

Acknowledgements

Sean Holler – IP considerations

Dr. Sidney Shapiro – Technology

Jason Innes – Business development

Others

17 of 17

Questions

Feedback on pivot from a 3D avatar to a voice-based AI agent?

EMR integration: pursue early partnerships with vendors, or maintain independence and prioritize a strong standalone data layer first?

Additional funding opportunities at this stage of development?