Why We Created Gravity Rail

Today we are telling the world about Gravity Rail.

Gravity Rail is a safe place for humans and AI to collaborate. We provide AI agents that can chat, call, text, email, collect data and connect to existing APIs and data sources, all while protecting patient data and escalating to a human where necessary.

We envision a world where everybody has access to world class healthcare, and where care providers spend their time engaged with patients rather than playing phone tag or dealing with repetitive busywork.

We have already started to have an impact. Gravity Rail's AI agents are already helping tens of thousands of real humans to take care of their kidneys, to get dementia support, to participate in clinical trials and so much more. As a result, our customers have been able to clear months-long backlogs of requests, to win more work from their sponsors, and to find patients needing an extra nudge to stay on track with their course of treatment.

A lot of smart people warned us not to focus on health care. They spoke of bureaucracy, entrenched incumbents, and tightly held networks of influence that determine winners and losers regardless of technical merit.

So I am here to answer that question. This post is not about the "what"; you can read more about that on our web site. This post is about the "why".

In October 2024, three profound but unrelated changes happened in my life, at the same time. Together, they pushed me to go all-in on Gravity Rail.

The first, and happiest, was getting married. None of what followed would have been possible without my partner Indigo. Anyone who has launched a startup knows the depth of resolve and patience it asks of a spouse.

Indigo is the reason that, at 46 years old and with two young children, I could even contemplate starting a new company. And as it happens, Scott Hoch — now Gravity Rail's CEO — was also at that wedding, and we started the conversation that night about what we might build together.

The second reason was more practical: I received a payout offer from my employer. It was not enough to retire on, but enough to last a year if I dramatically cut expenses. After 10 years at Automattic, including two as lead AI architect, I had dozens of ideas I wanted to pursue.

The third, and most pivotal, was my mother. Shortly after our wedding, Indigo and I flew to Australia to help out Mum while she recovered from a series of health crises.

Joan Walmsley is an indomitable Irish woman who has lived a life of physical fitness, self-sufficiency, and service to others, especially those less fortunate. Yet at 87 she found herself with low mobility, very poor vision and hearing and had great difficulty using a computer or smartphone. At the same time, her life had become an ever-more-complex regime of changing medications, appointments, home care providers, government agencies, and more.

What I learned while observing and helping coordinate my mother's care is that our health care system is in crisis, and many of the "solutions" we have tried are making things worse.

If you have ever cared for an ailing loved one, the following will seem depressingly familiar:

Phone numbers with interminable phone trees and hours-long waits to speak to a person who can help you
Having to repeat the same information over and over
Highly variable outcomes, no single person who knows the patient's whole situation
High staff turnover due to high workload and broken information systems
Patients and families pushed to buggy apps and web sites that are practically inaccessible to those with vision, hearing or cognitive impairment
An overwhelming daily burden of reminders, medications, measurements, callbacks, and logistics, with little or no tracking of adherence or outcomes, which can delay critical interventions
Long waiting lists and soaring costs for residential care, pushing more responsibility back onto families and fragmented home-care providers

And it's getting worse! Healthcare is getting more expensive, yet it's never been harder to get help when you need it.

So I decided to investigate.

The root cause is not mysterious. We are asking a strained and unstable workforce to coordinate increasingly complex care for a much older population. In the United States, every responding state in a 2024 KFF survey reported home-care workforce shortages, and 42 states specifically reported shortages in case managers. In Australia, the picture is just as stark: in home care, an estimated 29,000 directly employed nursing and personal care staff — 33% of the workforce — left in a single year, with 14,300 vacancies still open; in residential aged care, 40,100 staff, or 24%, left and 15,800 positions were vacant. This is not a temporary bottleneck. It is structural fragility in the human layer of the system.

That fragility collides with demographics. In Australia, the population aged 85 and over is projected to double to more than 1 million people by 2042, and people 85 and over already account for 54% of admissions to permanent residential aged care. Among Australians 65 and older living in permanent residential aged care, 37% had high care ratings across all three care-need domains, and more than half had dementia. In the United States, the 85-plus population is projected to grow from 6.5 million in 2022 to 13.7 million in 2040. Need rises sharply with age: about 8% of Americans over 85 live in nursing homes, compared with 1% of those aged 65–74, and someone turning 65 today has almost a 70% chance of needing some form of long-term care.

Families feel the consequences as delay, repetition, and cost. In Australia, more than 121,000 people — including my mother — were waiting for a Home Care Package at their approved level as of September 2025. For medium-priority approvals, estimated waits were 9–12 months for Levels 1, 2, and 3, and 12–15 months for Level 4. Earlier AIHW data showed median waits of roughly eight to nine months for people approved for home care. In the United States, the pressure shows up less as a single national waiting-list number and more as shrinking supply and rising prices: CMS-certified nursing facilities are down 6% since 2015, and a semi-private nursing-home room now costs roughly $112,000 to $115,000 a year nationally.

Even the workforce model is getting more brittle. In Australia's home-care system, the share of casual or fixed-term positions rose from 3% to 22% for nursing staff and from 8% to 40% for personal care workers between 2020 and 2023; in the Commonwealth Home Support Programme, casual/fixed-term roles rose from 4% to 52% for personal care workers. When staffing looks like that, continuity of care becomes an achievement rather than a baseline assumption. Context falls through the cracks. Families repeat themselves. Care workers spend more of their day reconstructing information than acting on it.

The first assistant I built was Seamus. Seamus has a phone number and speaks with a reassuring Irish accent. My mother can reach him by pressing a single button. He knows her world — her contacts, her routines, her limitations, the realities of her care — and he can work across the channels real people actually use: phone calls, voicemail, SMS, email, and the web. But helping just one person was no longer the goal. The deeper problem is that the systems supporting care coordinators, nurses, and home-care providers are still too fragmented, too brittle, and too labor-intensive for the scale of what is coming.

At the same time that this was happening with my mother in Australia, the same situation was faced by Scott in caring for his father back in Los Angeles. We both saw the same system in crisis, and the same opportunity for AI to help families, patients and clinicians to achieve new standards of health care.

As we explored more opportunities, we found that similar issues were faced by clinical trials. Clinical development is the largest cost center in bringing a drug to market, with Phases 1 through 3 accounting for about 68% of out-of-pocket development costs in one recent NCBI model. FDA guidance on decentralized trials is explicit: bringing trial-related activities into participants' homes can reduce travel burden and improve engagement, recruitment, and retention, and using local healthcare providers can help reach populations that traditional sites miss. But that flexibility only works if the operational backbone is strong — sponsors must plan for multi-source data flows, protocol discipline, and audit-ready data management. NIH recruitment guidance points in the same direction: choose the right channels for each community, whether that means face-to-face outreach, radio, newsletters, faith-based events, or local media, and use navigators and tracking systems to manage participant relationships over time. In other words, clinical trials are also a coordination problem — one with tighter compliance, documentation, and data-integrity requirements, but still fundamentally about reaching people where they are.

Across care management, home care, and clinical trials, we kept arriving at the same conclusion: the bottleneck was not compassion or expertise. It was coordination.

This is why we built Gravity Rail.

Not because health care is easy, but because it is hard. Because it is fragmented. Because too much of the burden falls on families, on care coordinators, on nurses, on front-desk staff, and on patients who are already exhausted. Because the people doing the most important work in the system are too often trapped inside broken workflows, brittle software, missed calls, duplicate forms, and impossible queues.

We believe AI can help with that.

Not by replacing the human beings at the center of care, but by giving them leverage. By helping a care manager keep context instead of losing it. By helping a family reach the right person at the right time. By helping a home-care provider coordinate across phone, SMS, email, and the web. By helping a clinical trial team maintain protocol discipline and data integrity while still meeting participants where they are. By making complex systems feel coherent again.

By liberating people from dehumanizing work, we help everyone to feel more human.

That is what Gravity Rail is for: to turn protocols into action, communication into continuity, and fragmented information into practical help. To make powerful, compliant AI systems that work in the real world humans actually inhabit — not the simplified world imagined by software vendors.

I started with my mother. Scott started with his father. But this is no longer just about our own families. It is about the millions of families, providers, coordinators, and patients facing the same pressures, and about the institutions trying to serve them with tools that were not built for the world we are entering. It is about the responsibility that those of us who can wield AI's power do so in the service of the greater good.

The aging wave is here. The workforce strain is here. The coordination burden is here. The question is not whether these systems need to change. The question is whether we are willing to build tools worthy of the people carrying them.

We are.

Today we are telling the world about Gravity Rail because we believe the future of care will belong to organizations that can combine human judgment, compassion, and accountability with AI systems that are reliable, secure, and deeply grounded in real outcomes.

That is the work in front of us. And we are just getting started.