Hi Reader,

We've been thinking a lot about how AI is changing what founders can build—and who they can build it for. One exciting trend we've noticed within the companies in our portfolio, is that AI is letting them stay lean, lower overhead, and reach communities that legacy systems previously left behind. We put together our thoughts into the essay below: AI as Infrastructure.

AI is making it possible to serve customers that legacy systems couldn't afford to reach before. That's not an efficiency play—it's infrastructure.

We'd love to hear what you think! If it resonates, please share it, and if you know any founders in health, wealth, and climate who are building for this world, we'd love to meet them!

Thanks for reading and onward!

Shruti Shah and Sydney Thomas,
Symphonic Capital

For decades, essential services have had a floor. If you couldn't afford the full-service option, you got a degraded version or nothing at all. It's just how the math worked: serving a patient costs $50 and you can only charge $30, so you don't serve them.

A patient at a small cardiology practice is on the books, but nobody has time to check if he's taking his meds. A loan applicant with a thin credit file gets auto-rejected before a human ever sees it. A gig worker can't prove income because the systems weren't built to recognize her. A rural clinic can't afford another full-time employee to handle patient calls. These aren't edge cases. They're the majority. But they’re priced out of systems that claim to be universal.

With AI, that's starting to change.

The Infrastructure Shift

Drop the cost to $15 and the structure shifts. But it's not just about cheaper. AI creates capabilities that didn't exist at any price point:

Personalization at population scale. AI can track thousands of patients simultaneously, catching when blood pressure trends up or medications aren't refilled, flagging problems before they become ER visits. That kind of individualized attention used to require one nurse per fifty patients. Now a small practice can offer the same continuous monitoring as a large health system.

24/7 availability in any language. A patient can text their care team at 2am in Spanish and get real guidance, not a recording. This didn't exist before, even for wealthy patients. The always-on, multilingual access that used to require round-the-clock staffing is now possible for a five-person clinic.

Handling complexity that humans couldn't do fairly. Evaluating a thin credit file requires weighing dozens of non-traditional signals. Humans either couldn't do it at scale or introduced bias. AI can assess patterns across data that humans never had time to consider, which means people who were auto-rejected before now get a real look.

Real-time dynamic matching. Matching workers to open shifts based on skills, location, past performance, and current demand, recalculating every few minutes, was economically impossible before. Now it's standard. A five-person staffing company can match talent as efficiently as national agencies with hundreds of employees.

These aren't incremental improvements. Infrastructure isn't about individual products. It's about what everyone can expect. Electricity didn't just help some people light their homes more cheaply. It reorganized the economy around the assumption that power is universal and always available. AI is doing something similar for essential services.

We're already investing in companies that prove it. Four of them are changing the economics in healthcare, gig work, and credit.

Welby Health

Seth Merritt spent twenty years at Blue Cross, Aetna, and CVS watching how value-based care was supposed to work versus how it actually did. The major players had no real incentive to keep patients at small practices healthy. The practices were too small to matter.

He started Welby as a side project, teaching himself to code and building a patient-facing app for type 2 diabetics. It grew into real customers and he quit his job in 2022. Today, nurses using Welby are 30% more productive, which means small physician practices can finally follow up with the heart disease and diabetes patients they used to skip. Welby's AI, Marcus, tracks patients between visits, catching blood pressure trending up or meds not being refilled, flagging problems before they become ER visits. Before AI, Welby wouldn't have been competitive. Forty-page RFPs that used to take five people two weeks now get done in a few hours.

Croux

Jennifer Ryan was running a restaurant in Birmingham during COVID with a newborn. When schools closed, working women on her payroll had to choose between paychecks and childcare. She saw exactly where the system broke and who it broke for. That sparked Croux, a staffing marketplace for hospitality. The founding team are all hospitality veterans who understand where the industry breaks.

Their CTO, Kenny Kung, previously built predictive credit modeling systems at BBVA. That same skillset, assessing trustworthiness with minimal data, now powers Croux's Trust Score. Traditional hiring screens on demographics or credentials, things people can't control and that often encode bias. Their score measures what's actually in someone's control: showing up on time, arriving in uniform, completing the work. It opens doors for people who couldn't have walked into a stadium and asked for a job. Five people are doing what used to require dozens. They've built AI agents for recruiting, matching talent to shifts, fraud detection, pricing optimization, and demand forecasting. They save clients five to ten hours per event and pass the cost savings back directly.

Conductiv

Gopal Swamy worked at a top-five US lender where he saw friction baked into every lending decision. He knew what was needed: not another point solution, but orchestration across the entire lending process.

Banks using Conductiv can now say yes to people they used to pass over. Lenders are 40% more efficient and increased approval rates by 47% with the same risk profile. Thin credit files that would've been auto-rejected now get a real look. They made a deliberate decision not to store personal information like names or SSNs. The AI finds patterns in other data, which means clearer decisions, less bias, and more people getting access to credit. The barriers weren't about risk. They were about cost.

Revel Ai

Christian Pean is an orthopedic trauma surgeon. He witnessed care inequities firsthand, one group of patients getting different standards of care than another. Most challenges patients face come down to access and communication with their clinical team, and as a surgeon, he felt powerless because so much of patient outcomes happen outside the OR. So he built Revel Ai.

His co-founder, Hadi Javid, brought deep NLP and LLM expertise. They started AI-first. The product handles patient communication: answering phones, booking appointments, synthesizing patient information, providing 24/7 access to care guidance via text, email, and voice in multiple languages. The AI delivers about half an FTE's worth of effort per organization, and NPS is above 90 for clinicians, staff, and patients. Patients who would've waited until morning, or until a problem became an emergency, now get answers immediately. Christian is the son of immigrants. His father came from Haiti, his mother from Mexico. His father is a primary care doctor who also built his own office complex. Watching his dad impact his community as both physician and entrepreneur shaped his path.

Why Proximity Matters

Infrastructure only works if people trust it and use it. You can build the most elegant solution in the world, but if it doesn't account for how people actually live, it sits unused. That's why proximity matters more than most investors realize.

When you've lived inside a broken system, you understand what doesn't show up in market research. You know which workarounds people actually use. You know what builds trust and what destroys it. You know where the official process says one thing but reality works differently.

That's not just about product-market fit. It's about building infrastructure that functions in the real world, not the one that exists in market research decks. When someone who's lived inside a broken system builds the replacement, they build something people will actually rely on.

What Comes Next

This is the beginning, not the end state. When small practices can compete with health systems on care coordination, what happens to referral patterns? When workers without traditional credentials can access formal employment at scale, what happens to labor mobility? When community banks can evaluate credit as sophisticatedly as national lenders, what happens to local capital formation?

We don't know yet. But we know infrastructure enables things that weren't possible before. The second-order effects of electricity weren't just "people have light at night." Factories could run 24/7. Cities could grow vertically. Entire industries that didn't exist became possible.

The second-order effects of AI infrastructure in essential services will be similar. We're not just making existing systems more efficient. We're changing what's structurally possible. Small climate adaptation firms will be able to serve frontline communities with the same sophistication that was previously only available to wealthy coastal cities. Workers will be able to build portable reputation that follows them across gigs and geographies.

On AI and Climate

We're not naive about AI's environmental costs. Data centers are driving up electricity prices. Communities are already paying for AI whether they know it or not.

That's exactly why this matters. AI is here. The infrastructure is being built whether we participate or not. The question is who it gets built for. If the only people shaping this technology are optimizing for efficiency and extraction, the communities footing the bill get left out. We'd rather be in the room, building something that actually serves them.

And this isn't either/or. We need AI infrastructure and we need investment in clean energy—more solar, more wind, more resilient grids. The software layer can help communities adapt faster, but it doesn't replace the hard work of building physical systems that last. Both/and.

What We're Building

At Symphonic, we back founders solving problems in health, wealth, and climate resilience. We've spent years watching these systems fail people, and we're doubling down on AI because we believe it can finally reach markets that never made sense before. That's not a trend we're chasing. It's what we're seeing our founders do.

We look for founders who are proximate to the problems they're solving. Not because proximity guarantees success, but because infrastructure that people don't trust or won't use isn't infrastructure—it's abandoned technology.

We look for founders who understand that AI isn't just about efficiency. It's about restructuring who gets served, who can compete, and what becomes possible.

And we look for founders whose business models don't treat universal access as a nice-to-have. We're not subsidizing access. We're investing in companies where access is the business model.

AI also changed what capital efficiency looks like. A founder used to need a big team and a big round to build anything meaningful. Now a small team with the right focus can outperform companies with ten times the headcount. That's why we invest the way we do. We come in after a company has raised an initial round, shipped something real, and started generating revenue. We lead $1–3 million rounds and stay price sensitive because we want founders to have options.

The companies that excite us stay lean, learn fast, and are using AI to get to clear proof points. You can see it in how they operate: five people doing the work of fifty, forty-page RFPs done in hours instead of weeks, AI agents handling tasks before there's budget for dedicated hires. Founders who automate before they hire. Founders who use AI internally, not just in the product they sell. Some investors see a modest raise or an extension and read it as a warning sign. We see founders who understand how this era actually works.

We're applying the same lens to climate. The communities getting hit hardest by climate change are the same ones legacy systems already underserve. Small teams using AI to serve those markets? That's exactly the pattern we're looking for.

The Infrastructure Exists. Now What?

The capabilities that used to require dedicated ML teams and significant upfront capital are now available through an API. A twelve-person team can serve populations that used to require fifty. Markets that were abandoned because the unit economics didn't work are starting to work now.

The question isn't whether this is possible anymore. It is. The question is who builds it, and who do they build it for?

Everyone says AI will increase inequality, and it could if we're not careful. We can see both futures from here—one where AI entrenches the same old patterns, one where it finally reaches people the old systems never did. There's a window right now to shape how this goes, and it won't stay open forever.

If you're a founder building for communities that legacy systems failed, we want to talk. If you're an investor who believes infrastructure should restructure power, not reinforce it, we want to talk.

The infrastructure exists. Let's build what comes next.