Can AI doulas help improve maternal health?

For pregnant women, the future is here—just in time for an urgent healthcare crisis. In the U.S., maternity care deserts are growing, often leaving communities completely without options. And despite the numerous warnings, nationwide maternal mortality data still paints a dire picture for pregnant women (especially rural women and women of color) who face preventable risks.

It’s a hard reality, but one where technology can be up to the task, helping humans improve outcomes and increase access in innovative ways. That means more remote tools and apps that help women monitor their vital signs, get information, and advocate for themselves. Also data-rich systems that allow health data to be shared securely with doctors, allowing for easier monitoring in-between visits. And now, AI-powered “doulas” that offer 24/7 support at a time when doctors’ visits are rushed and sometimes even nonexistent.

If you’re skeptical of AI, you’re definitely not alone. Across sectors, the rise of AI has come with plenty of promises, many of which have fallen short. But as someone who had her own pregnancy scare, which led me to develop a pregnancy app and AI doula with a human-centered approach, I’ve come to understand that maternal health is a challenge that AI may be uniquely suited to face. Not as a replacement for real doctors and doulas, but rather as a way to empower them and the people they serve—lowering costs, increasing their reach and effectiveness, and helping bridge a growing healthcare gap.

But first, the companies building these AI assistants have to do more than move fast and break things—they must earn trust. 

A helping hand

Obviously, AI doulas are not a cure-all when it comes to the complex issues pregnant women face as a result of our imperfect system. After all, AI still does not have the nuance to automatically navigate unwieldy situations where details matter.

To see what I mean, try experimenting with AI image generation tools and enter in “pregnant mother cradling her belly.” The moment you specify an ethnicity, you’ll notice a difference in image quality—African American and Native American images have the wrong number of fingers, mismatched eyes, or any number of other deficiencies in comparison to white and Asian ones. This suggests an error in the training data and an insufficiency across certain groups. If this is happening with basic image generation tools, imagine the impacts in clinical decision support tools.

That said, I’ve seen firsthand that AI is really good at culling through medical imaging, such as ultrasounds and MRIs, to make conclusions about the presence or absence of certain criteria or conditions. It is also really good at communicating confusing medical results in a way that everyone can understand.

That’s what makes AI doulas—aided by and leveraged alongside real doulas and medical professionals—a perfect application of the technology. They can remove many of the barriers of communication, information, and cost that currently plague the maternal health system. Curbing costs, in particular, can’t be underestimated. Despite the very important historical and cultural role that they fill, especially in BIPOC and rural communities, real doulas are still a sizable investment on behalf of an expecting family and are not consistently covered by insurance. AI can stand in for doulas after hours without increasing costs or acting as a replacement for the real thing. 

Then there are the instances when women have questions that might be uncomfortable or make them feel too awkward to ask a friend, member of their care team, or their partner. Or, since pregnancy can be lonely, people might not have a trusted person in their network to ask questions. Other times, distrust between women and their doctors—especially when it comes to disclosing mental health issues—results in important warning signs falling through the cracks. In these circumstances, the comfort of an AI doula can calm anxieties, give women control, and provide useful answers to help them advocate for the care they need. 

Combating skepticism

But if AI doulas are to have a positive impact on women’s lives (not just on health tech companies’ bottom lines), we must learn to trust it. That means it’s crucial that they’re developed in ways that strengthen what’s already working rather than promising to “disrupt” everything completely.

Disruption may seem like an exciting blueprint for innovation, but it’s an ineffective philosophy for building trust. So instead, conceptualize the technology as a support tool, not a substitute for doctors or the human connection and cultural wisdom doulas bring. Do the important work of giving women a sense of ownership in every AI feature, embodying a deep understanding of how the tech impacts not just pregnant women but the communities around them. This is true for all women, but especially those from minority communities facing mistrust born from longstanding maternal mortality inequities.

It’s not an easy task. For our personal virtual doula, Aria, we talked to hundreds of doulas over the course of a long development, involving birth and community members directly in the design, testing, and rollout so we were confident it reflected lived realities and priorities. We are also transparent about how any data on our app is used, stored, and protected and this helps to allay concerns, especially in communities with a legacy of medical mistrust.

Above all, AI must enhance the human qualities that make doulas and healthcare providers invaluable parts of the pregnancy journey. Care and comfort must not be sacrificed for efficiency—for pregnant women and their doulas, too. After all, being a doula is more than an occupation to be disrupted. It’s a calling that must be nurtured if they are to continue to play a crucial role in improving maternal health. With the right approach and eventual trust, families can leverage AI to navigate an evolving maternal health landscape with more confidence and continuity.