Women's health startup Millie developed an artificial intelligence agent specifically to support maternal health as its first foray into agentic AI.
The agent, called Maia, represents a personalized, "always-on" approach to women's health support and is just one part of the company's overall AI strategy, Anu Sharma, founder and CEO of Millie, said.
The AI assistant integrates into Millie's patient-facing mobile app, allowing patients to find answers quickly, handle appointments and get important reminders about their pregnancy care.
"We think of Maia as being a patient companion that is available to patients through the Millie app and has the ability to query all the data that is available on the back end," Sharma said in an interview with Fierce Healthcare.
"Today, we prioritize what we want patients to have less friction with, which is being able to get answers to basic questions that may be sitting in the app or in our handbook, or they may be waiting for our patient's navigators to answer nonclinical questions," she said.
With Maia, patients no longer need to wait for responses or read through hundreds of pages of guides and FAQs to get a response to common questions.
The company said it built the AI agent with patient safety, data privacy and clinical quality at its core. Maia customizes every interaction to each patient, remembering past conversations, learning communication preferences and adapting its tone and guidance over time based on each patient’s medical history with Millie, executives said.
When patients ask Maia a clinical question, the AI agent automatically escalates that request to the company's clinical team, a safeguard that Millie specifically built in, Sharma said.
"That also allows us to identify what types of clinical questions our patients might be facing, so that they can really think about—how do we prioritize questions like that appropriately to our clinical team for first-level triage. That's the capability we are building with more interaction data from patients," she said.
Millie is a California-based hybrid women’s health clinic offering both inpatient and virtual care services. The company, which launched in 2022, was originally focused on maternity care. Staffed by nurse-midwives and doulas, Millie provides care for low-risk pregnancies out of its clinic. Millie also partners with hospitals, where its providers support labor and deliver babies. The company operates two brick-and-mortar clinics—one in Berkeley, California, in partnership with Alta Bates Summit Medical Center, a Sutter Health-owned hospital, and a second location in San Francisco's South Bay in collaboration with Good Samaritan Hospital, part of HCA Healthcare.
In the past three years, the company has expanded to offer preconception counseling and gynecology services, along with perimenopause and menopause care. Millie picked up a $12 million series A funding round in February to help fuel its growth.
Millie aims to provide a new model of care that is "more complete, more proactive and more right-sized," Sharma said. "With our first clinic, we proved out all of that—we can build better care and we can do that with great outcomes, with better experiences and with a cost structure that is good for payers," she said.
The company plans to open between two and four new clinics next year with new health system partners, Sharma told Fierce Healthcare, and the company will expand beyond California. Millie has served 2,500 patients since its launch in 2022.
The startup's founding mission is to close persistent care gaps in women’s health by harnessing cutting-edge technology, and its AI strategy builds on that vision, Sharma noted.
Earlier this year, the company rebuilt its data and analytics platform from the ground up with AI in mind, Sharma said.
"If you really think about a company like ours, we have all kinds of data sitting in all kinds of buckets. We've got [electronic health record] data; we've got lab data, scheduling data, business data. As we prepare for an AI-forward world, the first thing we did was to actually unify all of these data sets into a single data set and then basically built a semantic layer over which there could be queries for us to be able to turn that into intelligence," Sharma said. "And then we built a front end over that, which is made up of AI agents, of which Maia is one, and that's our patient-facing agent."
The startup began its data rebuild in April. "We've been live on the ground since 2022, so we already had a lot to build from. What we didn't really have before was the capabilities of AI, and that hit the mainstream this year, for all intents and purposes. We said, 'Well, the time is right, and we should really be building for an AI world as we go forward.' So we basically rebuilt our entire data layer to be ready for a world that is going to be AI-powered," she noted.
“The healthcare system is full of valuable patient data that sits trapped in disconnected systems,” said Chase Schwalbach, senior vice president of product and technology at Millie. “Most legacy infrastructure was never designed for AI to access or understand that data. At Millie, we’ve built a centralized, AI-ready data platform that connects clinical, operational and patient-facing information in one place. This infrastructure is what has allowed us to launch Maia.”
The company's data platform rebuild gives it more speed and scalability in the world of AI as compared to healthcare companies using legacy tech systems, executives noted.
"If you look at a typical health system, there's not just the EHR, there's layers and layers of patchwork, so a lot of data sits in silos, and it's often difficult to talk to each other, and there's mountains of technical debt. It's much harder for them to retrofit AI on what is old architecture. For us, it was a lot easier to build this from the ground up to and have this structure going forward. In some ways, I feel like we had less complexity to 'undo' and then really build forward from," Sharma said.
Millie's AI agent can offer a level of personalization and context that legacy health systems can’t match, according to the startup's executives.
Millie's patient-facing app, which Sharma describes as "robust," already features structured content, integration with connected devices and the ability to do patient screenings between visits.
"We had that human interaction already kind of built into it and then we've essentially inserted Maia as an interface layer. Many health systems don't have that and users interact with them through a patient portal. We already had consumer content into which we could plug this into," she noted.
For clinicians, Millie is also using ambient clinical documentation and charting tools to extract and structure data that traditionally remains siloed in EMRs.
"We're creating a clinician-facing application layer," Sharma said.
Millie’s proprietary AI technology is built to detect and address risks earlier, automatically flag potential concerns based on patient and provider data, and focus clinical attention on urgent needs. With access to longitudinal data and real-time insights, Millie is also building a unique dataset that will deliver an increasingly more personalized model of care, executives said.
Sharma also sees the potential for agentic AI to assist clinicians with triage as a clinical use case.
"We certainly see an opportunity for there to be some level of stratification and prioritization, potentially as one potential application. I need to see a lot more before I can get comfortable around letting AI respond to patients on anything clinical, but I do think there's an opportunity for us to create an assistive layer for our clinicians to be able to filter and prioritize messages to the top of their inbox that may need their attention and get patients care in a more timely way and faster with a human in the loop," she said.
The company is also using AI to streamline clinical workflows and drive operational efficiency across the business.
The company’s new architecture includes automated data ingestion and a semantic data layer that enables agentic AI to reason over information in real time. With all of Millie’s data centralized and contextually linked, business users can now ask questions in plain English and instantly generate charts, spreadsheets and deep analyses, according to the company.
"This allows users, in real time, to see our clinical capacity, our appointment availability, any patients that are requiring follow-up, support tickets to how our patient churn data is trending," Sharma said, adding, "This tech-first approach is something that health systems have always aspired to and really had a hard time with, and we have already leapfrogged them, and then we're leapfrogging even further."
After opening its second brick-and-mortar clinic in April, Millie has seen increasing inbound interest from health systems "looking for next-generation solutions" for women's health that deliver growth with cost efficiency, Sharma said.
Health systems are facing OB-GYN shortages and a new generation of patients demanding a different kind of care experience, she noted.
"We have been responding to that interest in how we're structuring our health system partnerships going forward. The reason why they're coming to us is the 'build versus buy' decision for many around do they try to create something on their own that can be a response to the OB-GYN shortage that they're experiencing or do they choose to partner? I think for some, partnering is the answer," she noted.
"I think our data speaks for itself in terms of our outcomes and our patient experience scores, and now increasingly, with a technology-enabled approach, our cost structures mean we can do it better, we can do it faster," she added.