Two years ago, Partnership to End Addiction began exploring how AI could strengthen the human side of our work — starting with Google’s Generative AI Accelerator and a gift from the Stavros Niarchos Foundation (SNF). Looking back, what surprises me most isn’t how much more efficient we’ve become. It’s how much AI has improved the quality of our human support.
One of our signature programs pairs parents struggling with a child’s substance use with trained parent peer coaches who have walked a similar path. Preparing someone for those first conversations with a family in crisis has always been challenging.
In the past, a coach might have had one or two opportunities to practice before going live with their first family. Many told us they still didn’t feel ready. We hoped they’d grow with experience — but we couldn’t know with certainty that they had truly mastered the skills beforehand. So we reimagined that training experience.
Today, our coaches practice each core coaching skill through AI simulations. They receive immediate feedback after every simulation and can practice as many times as they want until they’ve demonstrated mastery. We don’t simply hope they’re ready—we know they’re ready because we can see it in their scores. And so can they. Instead of wondering whether they’re prepared, they begin their first coaching session with confidence.
That has changed the experience for everyone.
Families receive the strongest level of support from their very first coaching conversation. Just as importantly, our coaches don’t feel anxious, like they’ve been handed a manual and told, “Good luck.” They feel thoroughly supported and prepared. Which makes for a great volunteer experience.
The other transformation has been quality.
Historically, organizations like ours could review only a small sample of coaching sessions — perhaps 20% if they were fortunate. Today, our AI-powered continuous quality improvement system reviews every coaching interaction, every day. It flags opportunities for training or additional support almost immediately. That level of quality assurance simply wasn’t possible before.
What surprised me over the last two years is that the best use cases for AI haven’t been about replacing people. They’ve been about helping people do deeply human work better. That’s become our philosophy: we don’t ask, “Where can we use AI?” We ask, “Where can AI help people do their work better, more safely, or with greater confidence?”
That work also gave us the expertise and confidence to pursue bigger challenges, including an NIH grant to develop and rigorously evaluate an AI assistant for parents and other concerned family members. We’re not building it because AI is exciting. We’re building it because public health needs evidence about where AI can safely extend human expertise, where human support remains essential, and how the two can work together to reach more families.
I’m optimistic about what comes next. Nonprofits are underfunded, understaffed, and stretched thin. If AI can reduce administrative burden, improve quality, and help trusted organizations reach millions more people without sacrificing human connection, that’s a future worth building.
Emily Feinstein, J.D.
Executive Vice President, Partnership to End Addiction