Daniele Farnedi

Most scheduling APIs are built to serve clicks. AI agents need APIs that can negotiate. This week we went deep with a large healthcare network that’s designing a brand-new internal scheduling API specifically for voice automation. Not a PoC. A production-grade interface that will soon handle real patient traffic. What mattered most in that discussion wasn’t JSON structure or field naming. It was this: Where do you place the intelligence - in the API, or in the AI? A few very real, very practical lessons from that session: 1️⃣ Returning a single “best” slot kills conversation quality If the API only ever returns one pre-booked slot: - the AI can’t show alternatives, - can’t reflect patient trade-offs, - can’t negotiate. You end up with robotic flows like: “8:00 is unavailable → transferring to an agent.” When the API instead returns a small pool of available slots (not pre-booked yet), the AI can: - group options (“I see several openings between 10 and 13”), - adapt to rejections in real time, - iterate without bouncing the patient to a human. In production, this difference alone can decide whether a 13-minute complex scheduling call ends in success or failure. 2️⃣ Pre-booking should be surgical, not global Pre-booking everything the API returns feels “safe”… and quietly destroys flexibility. What scales better: - Fetch a wider set of available slots. - Pre-book only the specific slot(s) the AI actually reads out loud. That mirrors how real receptionists work. And it prevents the system from blocking inventory during long negotiations. 3️⃣ Time filters should be anchors, not cages “Morning / afternoon / evening” buckets seem helpful. Until a patient says: “Something around 16:00.” If your filter closes at 15:45, the AI is forced to say: “There’s nothing available.” If your filter is an open anchor (“search from 12:00 forward”), the AI can instead say: “I don’t see anything exactly at 16:00, but the closest is 16:20.” That single sentence is often the difference between: - conversion, and - escalation to a human. 4️⃣ APIs should stay dumb about policy - on purpose Cancellation rules, protection against critical specialist slots, double confirmations via SMS - all of that belongs in: - the AI layer, or - a domain orchestration service. Not hard-coded into the API. Because the moment clinical policy changes, you don’t want to wait for a new backend release to stay safe. ⸻ The pattern that keeps repeating across our deployments is simple: - Rigid API → fragile AI. - Flexible API → adaptable AI. If you’re designing a scheduling API today and expect AI agents to use it tomorrow, the real question isn’t: “Does it support appointment booking?” It’s: “Does it support negotiation?” That’s where real automation either works - or quietly collapses.

24

2 Comments

Good to see basys.ai featured in this market map on Utilization Management and other health plan/payer administration platforms. More importantly, it reinforces what we’ve believed from day one: UM is overdue for a rethink. At basys.ai, we’re building utilization management the way it should have been designed from the start. AI-native, clinically grounded, and built from first principles. Less administrative friction. More consistency. Better decisions. The broader healthcare market is starting to recognize that legacy workflows were never built for the complexity of modern care.

5

1 Comment

Wags Take: The HealthTech Operating Model, Ten Years Later. 10yrs ago a healthtech startup sold software. Today it sells outcomes. In 2015 the typical Series A team was 40 to 60 people. Big sales org, implementation services and integration engineers writing custom HL7 interfaces. 12-18 months from first call to first invoice. Per-user per-month subscriptions. The current model looks different from the inside. Teams are smaller, 10-20 people generating revenue that used to require a hundred. AI is doing the work headcount used to do. Pricing has also moved with it. Performance-based contracts are no longer fringe. Percent of recovered revenue. Percent of denials overturned. Per-claim fees. The vendor gets paid when the result lands, not when the contract signs. That changes the math on both sides. The buyer is not absorbing the risk. The vendor is. Which means the vendor has to know the product actually performs before pricing it that way. Distribution has shifted. The path to a health system used to be direct field sales. Today it is more often an embed. Through a clearinghouse. Through an EHR vendor. Through an RCM partner already inside the workflow. The new startup does not show up at the front door. It is delivered through a partner already in the building. The founder profile has shifted too. 10yrs ago the pairing was a clinician and a technologist. Today it is more often an operator and an AI engineer. A 2015 Series B healthtech burning four million a month was normal. A 2026 healthtech can reach ten million in ARR with twelve people. The second one did not used to be possible. Here's the wrinkle. Not all investors have caught up. Inside the Bay Area the new model is well understood. Outside it, many investors still evaluate healthtech startups against the old playbook. Per-seat ARR ramps. 36 month CAC payback. Sales team headcount as a proxy for go-to-market maturity. Traditional pure SaaS unit economics. A 12 person company doing 10MM in ARR through outcome-based pricing does not fit those rubrics. So it gets compared to companies that look like 2018 healthtech and asked why it does not have a thirty-person sales org. The founders who have had to explain their operating model twice in every meeting know what this looks like. A regional capital gap is forming. Companies built for the old model and the new one are no longer competing on the same terms. The buyers are starting to notice. The investors are next. #HealthTech #HealthcareAI #StartupFunding

4

don't become a healthtech founder because you're passionate about solving healthcare problems. become a healthtech founder because you have the only skills that matter: clinical grade delusion, the attention span of a goldfish and the ability to take hits like Homer Simpson

118

23 Comments

A healthtech founder ran a sell-side process with a well-known bank. The buyer was ready to do the deal...the deal still died. The finance team just couldn't defend the numbers under diligence. The founder ended up replacing the CFO and brought in an operator who's been through exits before. Multiple buyers are now circling back, and a deal should close. But what founders have to keep in mind is that bankers can create a process for you, but they can't fix a finance function that breaks under diligence. Sometimes the question isn't whether you're a viable, buyable business. It's whether your books are clean, your docs are organized, and if your financial story holds up when someone pressure-tests retention, margin, concentration, and unit economics. Before you hire a banker, stress-test your CFO.

24

7 Comments

New tech, who dis? Fifteen years designing healthtech experiences has taught me that the solutions change, but the problems don't. We've been lucky to work with clients long enough to see the full arc — from on-premise software to cloud, and now to AI-native products. The technology looks completely different each time. The underlying problems don't. It's a strange and fascinating thing, solving the same problems three times over with a completely different toolkit.

29

The best investors in Silicon Valley are all pushing AI-native services startups. Here's the truth behind building one to $612,000 ARR then failing: I co-founded Ply Health (YC S24), a credentialing service for behavioral health providers. In theory, we'd sell the high ticket outcome then collect the spread by using software to do the work. But the reality is trickier. 1/ A lot of work isn't automatable by AI (yet) AI couldn't navigate payer portals or fill PDF forms with enough accuracy, so we still needed a human in the loop. Even with AI supercharging every human, this still capped our output on headcount like a traditional services business. 2/ More of the value chain, more of the responsibility When you sell software, many failure points are absorbed by the company using your tool. When you sell the outcome, those now become your problem. Instead of a product making money while we slept, we spent many sleepless nights figuring out how to appease an uncooperative payer. 3/ Selling services is not selling software Like most startup founders, we were two young bright techies. I'd like to think you'd reasonably buy software from us. Would you buy credentialing from us? What if your alternatives are RCM companies with 100+ years of experience, processing millions of cases? Still, I'm bullish on AI native services. We just can't assume every service is a pile of cash waiting to be automated. You need two things: a critical look at what AI can do for that specific workflow, and founders who can actually get in the door. --- If you or someone you know works in an industry with lots of repetitive outsourced work, please reach out. I'd love to learn more.

75

13 Comments

That's exactly how I expected it would happen: you need the business skill as much as the AI skills to pull it off.

7

🚀 The past couple of days have been transformative at the intersection of digital health and policy. On May 29–30, we saw: • CMS doubling down on interoperability with fresh RFI momentum and the 1upHealth prior auth launch 🔁 • CHIME and HITLAB spotlighting the next gen of digital health leadership and regulation-ready innovation 🧠 • Ontario unlocking digital health identifiers via PHIPA reform, setting a new precedent for patient identity and access in Canada 🇨🇦 • Real-world AI deployments from Vancouver to Toronto proving that we’re beyond pilot projects—AI in healthcare is scaling, now 📈 The public and private sectors are finally speaking the same language: access, automation, and accountability. Let’s keep building where policy meets progress. #DigitalHealth #HealthTech #Interoperability #AIinHealthcare #PolicyAndInnovation

2

OpenAI just launched for healthcare. The technology isn't new - but what just changed is massive: Last week, OpenAI announced: • Optimized APIs for healthcare • ChatGPT Health for consumers. • The acquisition of Torch, an EHR startup. But what everyone's missing is that most of this already existed. • The APIs were already available. • The technology has been ready for months. So why does this announcement matter? 1. It validates the market. When a company like OpenAI publicly declares they're entering healthcare, that's a signal. When they make a healthcare acquisition, it's even clearer. Every healthcare investor, every clinic operator, every hospital system knows AI in healthcare is real and happening. 2. It educates the buyers. 90% of the clinic operators I talk to didn't know Open AI's APIs could already be set up to be HIPAA compliant. OpenAI just did the market education work that benefits everyone building in this space, including us at Entropy. 3. The implementation gap is still huge. OpenAI's announcement removes the knowledge barriers that kept many healthcare organizations from adopting AI. But here's what still breaks in the real world: What most healthcare organizations need is a system that learns how your clinic operates day to day. That adapts to change without engineers. That detects new providers and alerts the team. That knows who uses fax and who uses the portal. General-purpose AI - even healthcare-optimized AI - can't do that. That's what we're building at Entropy. OpenAI has validated the market. We're building what actually works in real clinics.

33

4 Comments

We launched an integrated shot tracker inside the Eden member portal, free for all members. Built around one clear idea, the feature converts everyday routines into measurable, clinically relevant signals. Quick logging, dose and site history and a clear burn-down of remaining medication let us capture real-world data to quantify adherence and iteratively optimize protocols. Instead of speculation we get empirical indicators that help clinicians make better, data-driven decisions in real time. For patients, it gives clearer control over their treatment, reduces the risk of dosing errors, and increases the likelihood of better health outcomes through consistent adherence. This is a step toward Healthcare 3.0, where personalization, seamless technology, and responsive care work together. When members can see exactly where they are in their treatment cycle, they can make better decisions with their providers, stay on track, and feel supported every step of the way.

23

🌍 Proof of Concept: Standardizing Oncology Care with FHIR. I'm excited to present Oncology Implementation Guide;a proof of concept developed to demonstrate how HL7® FHIR® standards can streamline oncology care. This guide provides a structured framework for capturing, managing, and exchanging oncology data—spanning prevention, diagnosis, treatment, and follow-up—enabling interoperability across healthcare systems. Explore the guide here: https://lnkd.in/ggMugMUj #Kenya #OncologyCare #FHIR #DigitalHealth #HealthTech #Interoperability #CancerCare #ProofOfConcept

11

5 Comments

Too many startups assume: 𝙩𝙝𝙚𝙧𝙚’𝙨 𝙖 𝘾𝙋𝙏 𝙘𝙤𝙙𝙚 → 𝙩𝙝𝙚𝙧𝙚𝙛𝙤𝙧𝙚 𝙩𝙝𝙚𝙧𝙚’𝙨 𝙖 𝙗𝙪𝙨𝙞𝙣𝙚𝙨𝙨. Reality is rarely that linear. Viability tends to unfold in fits and starts, shaped by payer coverage, provider workflows, and the real operational burden behind reliable billing. ⚙️📉 Inspired by Matt Kamen's excellent post on CPT adoption curves & the recent ACCESS discussion, we’re sharing our internal framework: 𝐐𝐮𝐞𝐬𝐭𝐢𝐨𝐧𝐬 𝐭𝐨 𝐀𝐬𝐤 𝐁𝐞𝐟𝐨𝐫𝐞 𝐁𝐞𝐭𝐭𝐢𝐧𝐠 𝐨𝐧 𝐚 𝐍𝐞𝐰 𝐁𝐢𝐥𝐥𝐢𝐧𝐠 𝐂𝐨𝐝𝐞 It goes deeper than payment policy — covering clinical and documentation requirements, enrollment and service mechanics, operational readiness, and the maturity of the CPT pathway. 🧠📋 If you’re evaluating a CPT-driven business model — or working with someone who is — comment “𝘾𝙋𝙏” below + DM me & I’ll share the document. 📄➡️ #startups #HealthcarePolicy #HealthTech #CMMI #ACCESSModel #Medicare #DigitalHealth

42

64 Comments

I don't fully trust GPT or any AI to work, but it is a good tool to rephrase some complex concepts. Always keep in mynd AI is not a "thinker" but a huge database with abilities to analyze information. However, at this point, it cannot be used as a source of information; it is still a tool.

4

📢 New in BHTY: Health Data Marketplaces: Why They Matter—and Why They’re So Hard to Build - AI’s explosive growth has made one thing clear: healthcare innovation depends on access to trusted, high-quality data at scale. Yet despite the promise, true health data marketplaces remain frustratingly out of reach. In this special ConV2X Decentralized Health Podcast, leading experts explore the realities behind the hype—and outline a credible path forward: Featuring insights from: - Francisco Curbera, PhD, SVP Engineering & Research, Accrete.ai - Shahram Ebadollahi, PhD, MBA, Founder & Principal, Nav.AI - Olivier Elemento, PhD, Director, Englander Institute for Precision Medicine, Weill Cornell Medicine The discussion tackles critical questions shaping the future of AI-driven healthcare, including: ✔️What defines a viable health data marketplace—and what features are non-negotiable ✔️How data value and incentives can (or cannot) be aligned across stakeholders ✔️Whether we’re facing a Prisoner’s Dilemma in today’s health data ecosystem ✔️How US and EU regulatory frameworks both enable and constrain progress ✔️What capital markets and VCs are actually funding—and what’s missing ✔️ The next critical step to unlocking permission-based health data access For those working at the intersection of AI, healthcare, policy, and decentralized innovation, this conversation is essential listening. Read / listen here: https://lnkd.in/e9XMARY5 #HealthData #HealthDataMarketplace #DecentralizedHealth #HealthcareAI #AIinHealthcare #DigitalHealth #HealthDataIncentives #PermissionBasedData #HealthDataGovernance #Web3Health #BlockchainInHealthcare

9

What kind of mindset does it take to go from building data systems at Palantir, to leading engineering at Komodo Health, to becoming CEO of Particle Health – and suing one of healthcare’s largest incumbents? Jason Prestinario is a First Builder through and through. As CEO of Particle Health, his mission is to fix patient data interoperability, so your primary care doctor, specialist, and surgeon can all make decisions using the same complete health record. What a concept, right? And yet, it’s far from reality. That’s why Jason and the team at Particle are now suing Epic, one of the industry’s largest incumbents, to push for real change. We talked about how he approaches challenges of this scale, and what keeps him grounded in the process. We sat down with Jason to talk about this and much more. Hope you'll enjoy this spicy and insightful episode! Link in comments 🌶️ 👇

20

1 Comment

“The thing I am most proud of is that because we built the entire tech stack that we own, from the member app, the company, the client dashboard, to electronic medical records systems and other internal tooling.   We enable providers to spend more time with the patient.” How much longer? 3x longer than the average primary care visit.   This week on the Digital Health Heavyweights Podcast I sit down with Joseph Kitonga, CEO and founder of Vitable Health. Joseph shares his journey from a family background in caregiving to creating a health plan that addresses the needs of underserved workers. He discusses the inspiration behind Vitable, his experience with the Thiel Fellowship, the prestigious Y Combinator,and the unique aspects of Vitable's direct primary care model.   “We are growing extremely quickly” The conversation also covers the challenges of hybrid care models, the importance of empathy in healthcare, and the company's growth trajectory. Joseph emphasizes the need for accessible healthcare and shares insights on building a successful startup in the healthcare space.   How is Vitable impacting clients? An average savings of about 12%   Takeaways ✨ Joseph's family background in caregiving inspired him to create Vitable Health. 💙 Vitable Health aims to provide affordable healthcare for underserved workers. 🎓 The Thiel Fellowship provided Joseph with the opportunity to focus on his startup. 🚪 Vitable's model reduces barriers to accessing primary care services. 📈 The company has about 100,000 members and is growing rapidly. 💻 Vitable's approach integrates in-home and virtual care effectively. 🤝 Building empathy through direct interaction with clients is crucial for Vitable. 🗣️ Joseph emphasizes the importance of talking to users to build a successful product. 🚀 The Y Combinator experience was pivotal for Vitable's growth. 🧘 Joseph practices mindfulness and reading to manage stress and maintain focus. Check out the episode and be sure to like comment and subsc https://lnkd.in/dVAzHMQZ

19