The pitch was simple, maybe deceptively so: What if artificial intelligence didn't just read medical scans, but actually fixed the problems it found?
That's the bet Bunkerhill Health made when it closed a $55 million Series B round this July, according to company disclosures. The San Francisco startup—backed now by Khosla Ventures and Sequoia Capital—has moved beyond the typical healthcare AI playbook of flagging abnormalities and leaving humans to sort out what happens next. Instead, its Carebricks platform deploys what the company calls "AI agents" across hospital systems, handling everything from cardiovascular screening alerts to prior authorization paperwork to billing documentation prompts. At the University of Texas Medical Branch, more than 20 of these agents are now running in production, each automating tasks that once required radiologists, specialists, care coordinators, and billing staff to manually connect the dots.
Whether that ambition will scale—or whether hospitals will balk at letting software make clinical and operational decisions across so many domains—remains an open question. But the funding, at least, suggests investors think Bunkerhill is onto something.
Khosla led the round, with participation from Sequoia, Felicis, Optum Ventures, and Y Combinator. The $55 million total includes both the recent Series B and earlier capital, according to Fortune's reporting. The Series B itself raised $25 million, following a $6.5 million seed round led by Sequoia in 2023. Bunkerhill says the money will fuel expansion across additional health systems and add new agent capabilities to its platform.
When Incidental Findings Don't Get Follow-Up
Carebricks was announced in February 2025 with a problem that radiologists know well: routine CT scans often reveal incidental findings that matter clinically but get lost in the shuffle. Coronary artery calcification. Aortic valve disease. Bone density issues. These show up in radiology reports, get mentioned in a line or two, and then... nothing. No referral. No follow-up appointment. The patient goes home, unaware their scan suggested they might benefit from a statin or a cardiology consult.
Bunkerhill built FDA-cleared algorithms to quantify those findings automatically. Then it wrapped them in software that could do more than generate a number. The system reads EHR data, applies clinical guidelines, prioritizes which patients need outreach most urgently, drafts referral orders, and sends notifications to the right clinicians. What emerges is software that doesn't wait for a human to interpret the data—it acts on it.
By mid-2026, the platform had evolved well beyond cardiovascular screening. Carebricks now handles referral triage, assembles prior authorization packets, generates clinical documentation improvement queries, manages preventive screening outreach, and interprets lab results. At UTMB, where the system was deployed across the organization in January 2026, the health system deployed agents across seven clinical domains. A KLAS Research case study published in March reported that UTMB achieved "100% automation" of its agentic workflows—meaning the agents complete tasks end-to-end without manual intervention, at least in most cases.
That's the promise, anyway.
What It Looks Like in Practice
The UTMB deployment offers a glimpse into how this works on the ground. One early agent flagged 11.7% of routine chest CT patients as having unaddressed, clinically significant coronary artery calcification—patients who needed statin therapy or cardiology referrals but hadn't received them. Another agent focused on clinical documentation improvement, capturing what Bunkerhill described as "hundreds of thousands" in additional revenue during its first month by prompting physicians to document diagnoses more precisely for billing purposes.
McLaren Health Care in Michigan launched a cardiovascular screening program in February 2026 using Carebricks to analyze a year's worth of prior chest CTs for coronary and aortic valve calcium. McLaren framed the initiative as a "first-of-its-kind" AI-powered screening effort in the state, relying on FDA-cleared algorithms to identify high-risk patients who might benefit from preventive cardiology interventions. The health system is among five in the U.S. deploying Bunkerhill's inflamed aortic valve calcium detection, according to company materials.
Bunkerhill also lists Cleveland Clinic and Intermountain Health as customers in its Series B announcement, though these deployments have not been independently confirmed in mainstream press coverage outside of company sources.
The Reimbursement Tailwind
Part of what's making broader adoption possible is money—specifically, a new reimbursement code. In early 2026, the Centers for Medicare & Medicaid Services established HCPCS code G0680 for AI-based coronary artery and aortic valve calcium analysis on chest CT scans. The payment, roughly $15.50 per scan under the Outpatient Prospective Payment System, gives hospitals a way to offset the cost of running opportunistic screening programs at scale.
It's not a windfall, but it helps.
FierceHealthcare reported in May that Bunkerhill held nine FDA-cleared clinical algorithms at that point, covering coronary artery calcium, aortic valve calcium, mitral annular calcification, abdominal aortic quantification, and bone mineral density assessment. The company has been steadily adding clearances since 2025, with its most recent 510(k) decision dated March 6, 2026, according to FDA records.
The reimbursement pathway matters because the economics of opportunistic screening can spiral quickly. STAT News, citing Bunkerhill CEO Nishith Khandwala, reported in April that at one health system, 63,000 of 120,000 scanned patients had coronary calcium scores above 100—a threshold that typically warrants intervention. Contacting all of them would overwhelm care teams. The platform's triage and prioritization logic—which layers in lab results, medication history, and prior cardiology visits—helps hospitals focus outreach on patients most likely to benefit, rather than triggering an avalanche of unnecessary follow-ups.
The Technical Layer
Carebricks operates as a layer atop existing EHR systems, pulling in imaging reports, lab values, clinical notes, and structured data in real time. The platform launched a conversational interface, Carebricks Chat, in August 2025, offering clinicians natural-language access to the full patient record with citations back to source documents. A "Context Control" feature, released two weeks later, lets hospitals tune how much historical data the agents consider when generating outputs—an accuracy-versus-latency trade-off that varies by use case.
The company markets the platform as "AI Agents for Modern Health Systems" and positions it as something of a build-your-own toolkit. Hospitals can configure agents for their specific workflows, from flagging sepsis risk based on lab trends to assembling prior authorization packets for high-cost imaging orders. The software holds SOC 2 Type II, HIPAA, and ISO 27001 certifications, according to Bunkerhill's website, though the company doesn't disclose when those certifications were issued.
The Investors' Bet
Khosla Ventures led the Series B, with managing director Kristina Shen describing Bunkerhill's approach as a shift from "AI as a tool" to "AI as a workforce" in a statement. Sequoia, which backed the company's 2023 seed round, emphasized the speed of UTMB's expansion from one pilot agent to more than 20 in production. The firm's investment blog, published July 16, framed Carebricks as an agentic platform serving "all health system functions," not just radiology or a single clinical specialty.
Khandwala and co-founder David Eng started Bunkerhill after working at Stanford's Center for Artificial Intelligence in Medicine and Imaging. Fortune reported that a personal experience—Khandwala's father had a routine chest CT that revealed coronary calcium but no follow-up—shaped the company's focus on closing care gaps revealed by incidental findings. The startup went through Y Combinator in 2019, raised its seed round led by Sequoia in 2023, and now employs somewhere between 11 and 50 people, according to LinkedIn data from mid-2026.
A Crowded, Fast-Moving Field
Bunkerhill is hardly alone in the race to deploy AI agents in healthcare. Innovaccer launched its Gravity platform in May 2025, positioning it as an agent orchestration layer with a unified data foundation for population health and operational workflows. Salesforce introduced Agentforce Health in March 2026, offering pre-configured agents for payer and provider service operations with FHIR integration. Oracle debuted a Clinical AI Agent at HIMSS in March, targeting emergency and inpatient physicians with note drafting and order suggestions embedded directly in the EHR.
Patient access platforms like Hyro are building agentic systems for contact centers, while Hippocratic AI is deploying voice-based agents for patient outreach and nurse support. The broader market momentum was visible in an Axios report from July, which highlighted CVS Health's pursuit of an "AI front door" strategy with Google—a signal that payers and retail health players are betting on agent-based interfaces as the next competitive battleground.
What sets Bunkerhill apart, at least in its current deployments, is the focus on back-end clinical and operational workflows rather than patient-facing interfaces. The company's agents sit inside the EHR, reason over longitudinal data, and execute tasks like referral creation or documentation prompts that traditionally required human coordination across multiple departments. Whether that proves more valuable than chatbots or voice agents is a question health systems are testing in real time.
The Harder Questions Ahead
The funding gives Bunkerhill runway to find out. With reimbursement in place, FDA clearances accumulating, and early deployments showing operational impact, the company has a window to prove that agentic automation can scale beyond a handful of pioneering health systems.
But the harder question—perhaps the one that will determine whether Bunkerhill becomes a lasting force or a cautionary tale—is whether hospitals will ultimately trust dozens of agents to make decisions across clinical domains. The technology may be ready. Whether the industry is ready for it remains to be seen.
For now, at least, the investors are placing their bets.
