Narration script
iECHO + AI: Scaling trusted human expertise, responsibly
About 8 minutes at a natural speaking pace
-
-
Opening
Welcome. In the next few minutes, we'll show you how we're bringing AI to iECHO, and why the trusted human network at the heart of the ECHO Model makes it different.
-
Step 1
ECHO works because of people. We've always used technology, but it's not what makes us unique. What makes ECHO work is the lasting connection between people: trusted, peer-to-peer learning communities that reduce the isolation frontline workers feel. Frontline workers share what really happens in the field, while experts share best practices to address those local challenges.
-
Step 2
AI is a powerful and highly scalable technology for sharing knowledge. But on its own, AI has no context about lived experience and what goes on in communities day-to-day.
-
Step 3
Together, they're transformational. When the learnings from peer-to-peer interactions are infused into AI, we can create impact at a scale neither could reach alone. Here's how it works.
-
Step 4
It starts with a collaborative virtual community, where specialists and local providers learn from each other's cases.
-
Step 5
Every session, case discussion and conversation on iECHO creates new knowledge.
-
Step 6
AI then securely harnesses these insights and summarizes them, with data privacy and governance built in.
-
Step 7
Those insights flow back to every participant, in their own language, and spark the next conversation.
-
Step 8
And what's most critical is that AI augments human connection and never replaces it.
-
-
-
Opening
For learners, iECHO + AI means the session never really ends. After each session, they get a short recap of the key points, in their own language, to read or listen to on the phone they already use.
-
Step 1
iECHO recommends programs and sessions matched to their goals and profession.
-
Step 2
Between sessions, they can practice on simulated cases built from expert-curated material.
-
Step 3
And when they have a question, they get an answer grounded in their own program's sessions, with the sources and the experts who said it.
-
-
-
Opening
Here's what makes this different. Ask a general AI chatbot what to watch for in their patients this month, and a health worker gets textbook guidance from the public internet, frozen at its training cut-off. It has no idea what's happening in their community.
-
Step 1
Ask iECHO, and they see what peers across the network are seeing right now: a rise in dengue-like fevers after flooding, raised in dozens of sessions across hubs in their region, what other health workers are trying, and an expert's review.
-
Step 2
These insights are current, local, trusted and private. They come from the trusted human network, and they keep evolving with it. No general AI model can offer that.
-
-
-
Opening
This brings us to the heart of our vision: the ECHO Listening Network. With consent, it listens to sessions, case discussions and questions across the network. It transcribes and translates them, de-identifies them at the source, and uses AI agents to find expert insights, questions and emerging needs. Subject matter experts validate the findings.
-
Step 1
Those insights build a living knowledge graph, connected by geography, hub, program, disease area and expert, and traceable to the sessions they came from.
-
Step 2
Now anyone with the right access can ask the network a question. Search for fever clusters in western India over the past 30 days, and an emerging signal lights up: dengue-like fevers after flooding, raised in 42 sessions across 9 hubs in two states, and validated by a hub expert.
-
Step 3
That gives ministries and public health agencies early signals before they show up in reports, gives learners insights from their peers in their own language, and gives program designers, funders and researchers evidence for their decisions.
-
-
-
Opening
The Listening Network powers a team of AI agents, one for every ECHO practitioner.
-
Step 1
For participants: a session assistant, a learning advisor, a practice coach and a support agent in more than 25 languages.
-
Step 2
For experts and hubs: an expert copilot that prepares sessions and reviews insights, an analyst agent that writes their reports, a program designer and a privacy guardian that catches patient details before anyone sees them.
-
Step 3
And for ministries, funders and researchers: listening agents that surface emerging needs, research synthesis across more than 900 peer-reviewed ECHO publications, impact reporting on demand and policy briefs grounded in the network.
-
Step 4
Every agent draws on the same knowledge graph, under the same guardrails, with people in the loop.
-
-
-
Opening
How we build matters as much as what we build. Human expertise and validation stay at the center, and these principles govern every AI feature on iECHO: open, interoperable public goods, so the community is never locked into one vendor; diverse, representative data grounded in local expertise; multimodal and multilingual by default; human validation and expert curation, with people accountable for decisions; and a trust and safety architecture that de-identifies data, so AI can scale without exposing anyone.
-
Step 1
And we make clear commitments to our partners. No partner data is used to train AI models. Partners can opt out of AI features at any time. We're open about where AI is used and what it does. And we roll it out gradually, together with our partners.
-
-
-
Opening
Here's how it's built. It starts with the iECHO platform itself: the iECHO app on the web, mobile and WhatsApp, open APIs for every capability, and the data on programs, sessions, people and outcomes.
-
Step 1
Around it, we add an agent harness, built on the same patterns as leading open agent frameworks. It's a reusable runtime: an agent loop that plans, acts, observes and learns, using iECHO's own APIs as its tools, reusable skills loaded on demand, memory of what each learner and program needs, and a scheduler that sends recaps and reports on its own.
-
Step 2
The harness powers domain agents: a session summarizer, a learning advisor, a practice coach, an analyst, a listening agent and a support agent. Each is the same runtime with different skills.
-
Step 3
The knowledge graph grounds every answer, connecting geography, hubs, programs and disease areas with expert-validated insights, a vector index for semantic search and more than 900 ECHO publications.
-
Step 4
An LLM gateway powers it all, routing each request to the right model, with caching and failover: Gemini, Claude, open-weight models hosted in country, and models for speech and embeddings.
-
Step 5
And trust and safety wraps the whole AI layer: guardrails, de-identification, evaluations, human review and a full audit trail.
-
-
-
Opening
There's an even bigger opportunity. AI is changing medicine faster than most health systems can adapt. Tools arrive before there's local guidance on using them well. Implementation is the hard part: validating tools, fitting them into workflows, training staff and governing their use. And the gap is widest where resources are thinnest.
-
Step 1
The ECHO Model was built for exactly this kind of challenge. An ECHO Program on AI in health care would put experts at the hub, connected to clinicians, nurses and health workers, hospital leaders, ministries, health IT teams and medical educators, all learning from each other's real experience of putting AI to work.
-
Step 2
Together, they'd work through what matters most: evaluating AI tools, fitting them into clinical workflows, governance and regulation, bias and equity, privacy and patient safety, and AI literacy for the whole workforce.
-
-
-
Opening
iECHO already connects more than 700,000 people through more than 1,200 hubs in more than 70 countries. With responsible AI, it can bring trusted, local expertise to every health worker, everywhere, and transform health systems at scale. We invite you to join us and help build it together. Thank you.
-