Jonathan Tushman
Boston, Massachusetts, United States
2K followers
500+ connections
Activity
2K followers
-
Jonathan Tushman posted thisHad a great conversation with Pat Brans for CIO Online on AI accountability and governance — link in comments. The essence of it: in this new AI world, you need people pushing it forward and people keeping is safe. The value is in that tension. At Hi Marley we've tried to design for the tension rather than eliminate it: - AI Ops owns how we build and run AI internally. - AI in the product belongs to the CTO and product leadership. - Compliance and legal act as independent checks and balances. The point isn't avoid the conflict. It's to institutionalize and lean into it. Thanks Pat and the CIO team for the conversation. #ResponsibleAI #AIGovernance #CAIO #InsurTech #AILeadership
-
Jonathan Tushman shared thisThe number two heaviest Claude user at Hi Marley last week wasn't an engineer. It was someone on our People team. Working on a PowerPoint. Honestly — that's great. We *want* non-engineers leaning hard on AI. But when we dug into why her token count was so high, the answer wasn't ambition. It was the file format. Quick math. Same one-page document — one heading, four paragraphs — in three formats: • Markdown: 362 bytes • HTML: 508 bytes • .docx: 36,840 bytes Roughly 100× larger for the same content. And the actual text inside that .docx? About 0.27% of the unzipped contents. The other 99.7% is machinery — theme files, style definitions you're not using, font tables, an embedded thumbnail you didn't ask for. Every time someone asks Claude to edit a Word doc, the model has to chew through all of that. Every turn. Every byte. That's where the tokens go. The deeper point: when AI generates output, its native languages are Markdown and HTML. Both are plain text, lightweight, infinitely expressive, diffable, version-controllable. Asking it to write a Word document is asking it to translate its own thinking into a 6,000-page-spec foreign format. (The OOXML spec really is ~6,000 pages. Markdown's spec fits on one web page.) But there's a second problem that's bigger than the file size. Word and PowerPoint have a *direction*. Word is vertical — top to bottom, like a scroll. PowerPoint is horizontal — left to right, like a strip of film. Both one-way. Information in 2026 isn't one-directional. It's networked. It branches. It links. It updates. It responds to what the reader knows. HTML was literally designed for this. Word and PowerPoint were designed for paper. We've been swapping decks and memos for single-page HTML artifacts at Hi Marley for months — internal briefs, planning docs, decision memos, executive readouts. Faster decisions. Denser information. No more waiting for the meeting where someone clicks through 47 slides. Anthropic, OpenAI, and the big AI vendors all know everyone uses Word and PowerPoint, so their integrations shoehorn AI back into those tools. Make slides faster. Generate paragraphs faster. The end state of that path: more PowerPoints, faster. Bigger Word docs, faster. Same one-directional artifacts — just in greater volume. It is time to unshackle ourselves from these file formats. Try it this week. Next time you're about to ask Claude for a Word doc or a deck, ask for a single-page HTML document instead. See what comes out the other side. It's lighter than a stone tablet. I promise. Full piece (covers collaboration)↓ https://lnkd.in/e7YJ6tpV
-
Jonathan Tushman shared this"Dad, I ran out of credits." My son called me last night. He's been using Claude Code to build things. Writing software. Running agents in parallel. He figured out git worktrees on his own and texted me about it — awesome. I had the privilege of saying: "Cool. Let me get you a Max plan so you're unblocked." Many families can't. Over the last few years I've toured maybe thirty campuses. The script is always the same. They walk you to the library: "If we don't have the book, we ship it from another library in the network." Then the maker space: "Students can 3D print anything they want." I'm standing there thinking — you are missing the point entirely. Not one school mentioned anything about AI. – ZERO. Knowledge in 2026 isn't hardcover books, no matter how romantic they are. And making isn't 3D printing. The knowledge that matters is access to intelligence. To tokens. The making that matters is what happens when a student opens Claude Code and ships something in an afternoon. I pay an obscene amount of tuition (don't even ask). And I still pull out my credit card to buy my kids the tools that actually matter. Something is wrong. Three simultaneous failures: * Universities should fund AI access the way they fund lab equipment. * AI companies haven't built the institutional access models education demands. * And public policy hasn't caught up to the reality that AI access is becoming as fundamental as Wi-Fi. This stopped being abstract when I thought about hiring. Claude Code has a feature that shows your usage insights (/insights). I was excited about it as a hiring signal. If a candidate showed me their stats, that would tell me a lot about whether they get this moment. Then I caught myself. If I make AI fluency a hiring requirement, I'm penalizing every talented person whose family couldn't afford the tools. That's a wealth filter wearing the clothes of a skills assessment. We need a Carnegie moment for AI. Andrew Carnegie recognized something similar 150 years ago. Peoples potential was unlocked when given access to knowledge. Back then it was books, so he then built 2,500 libraries. Universities already negotiate consortium licenses for journals and databases. The same model works here. Every enrolled student gets a meaningful monthly token allocation, the way they get library access, lab time and Wi-Fi. Somewhere right now, a student is hitting the same wall my son hit last night. My son got unblocked with a credit card. Another student might not. One kid keeps building. The other stops. That’s the gap we need to close. #AI #EdTech #HigherEducation #DigitalDivide #FutureOfLearning
-
Jonathan Tushman shared thisMassachusetts Institute of Technology's AI Summit is starting today! And I'll be joining Lily Lyman Jordan Hayashi and Cecilia Liu on Friday for a conversation on what it take to successfully build and scale #AI at startups / scaleups Excited to connect with and learn from others pushing the boundaries of what's next in AI. Learn more about the event: https://lnkd.in/ejbXwCy2
-
Jonathan Tushman shared thisClaude called in sick. Here’s what I learned. For the past 36 hours, I haven’t been able to log into Claude Code. Most of my team hasn’t either. We’ve been blocked by authentication issues, and it created a very real disruption to day-to-day work. A few things became obvious very quickly. First, AI tools have already crossed the line from “interesting hobby” to mission-critical infrastructure. This wasn’t just an engineering team inconvenience, people across the organization felt it. When a tool disappears and it impacts how multiple teams work, you realize it’s gone from experimental to operational. Second, switching costs are surprisingly low. After 15 to 30 minutes of waiting, I moved over to OpenAI Codex and was productive again pretty quickly. The workflow was similar enough that I could keep moving. But I still miss working with Claude. It feels like working with two excellent engineers who just have different personalities, strengths, and instincts. I know Claude really well. I know its tone, its edges, how it responds, and how to get the best out of it. That familiarity matters and it’s still my preferred way of working. But using Codex has been genuinely useful too, especially as a second set of eyes on projects I originally built with Claude Code. I’ve gotten some interesting refactoring ideas out of it already. Third, the uptime conversation is going to get a lot more important. When you look at the status page and see an uptime number that would feel unacceptable in most enterprise software categories, it raises a question, “ What happens when your product increasingly depends on cutting-edge AI providers that are still going through hypergrowth pains?” Our customers will continue to expect the uptime, resilience, reliability, and SLAs they count on from Hi Marley, and they should. That bar doesn’t move because an upstream model provider has a rough week. It’s our responsibility to architect for that reality, and we do. Model choice, failover strategy, vendor diversification, and hosting decisions aren’t side conversations. They’re part of how we deliver enterprise-grade software. I have a lot of empathy for the Anthropic team. Runaway success creates operational strain, and this is what scaling pressure looks like in public. But as builders, especially those of us serving enterprise customers, we need to design for this reality. Claude, get well soon. I’m looking forward to having you back.
-
Jonathan Tushman shared thisYes, and… Every week my LinkedIn feed fills with people bragging about how many tokens their engineers burn, or stating don't worry about the tokens this is what productivity is now. Last week, the CEO of Nvidia said your $500K engineer should be spending $250K a year on tokens. My first reaction was honestly: Am I missing something? I'm on a Claude Ultra account, coding every single day, shipping at a pace I've never shipped before — and I'm nowhere near those numbers. I like never hit my quota. So maybe I'm missing something. Who remembers Heroku and the Dyno? ~15 years ago, Heroku democratized cloud computing. You could spin up a Ruby on Rails app and deploy it in minutes. It was revolutionary. They had this concept called a Dyno — a unit of compute with fixed constraints. The smallest one gave you 512MB of RAM. When you hit your first out-of-memory error, some people saw that as a bug. It was a feature. Those constraints taught you how to build proper database indexes. How to paginate. How to be intentional about the context you load into a page. Those guardrails forced you to build solutions that were infinitely scalable and cost-efficient from day one. The skills I learned from working within those constraints shaped how I think about architecture to this day. I think the every growing context windows are going to be a crutch and can lead to bad inefficient engineering. So here's what I think is missing from the "look how many tokens we burn" conversation: the word Pragmatic. Listen, I'm all for using AI to build at speed and at scale. This isn't an either/or. It's a yes, AND: Yes, use AI aggressively — AND think about token efficiency. Yes, empower your engineers — AND put patterns and constraints in place. Yes, move fast — AND be thoughtful about when you release those constraints. Constraints aren't the enemy of productivity. They're the foundation of it. #AI #SoftwareEngineering #TechLeadership #AIEngineering #Pragmatic
-
Jonathan Tushman shared thisExcited to share my conversation with Stephanie Behnke on the debut episode of Conversations that Connect. We dug into a question I think about every day: How do we use AI to make the claims experience *more* human, not less? The short version: adjusters spend too much time on repetitive tasks and not enough time on the moments that actually matter to policyholders — empathy, clarity, reassurance. AI can flip that ratio. A few things we explored: → Compliance as a Superpower → The mission-driven advantage of insurance → Creating space for group-up innovation Would love to hear what resonates with you — give it a watch and let me know. 🎬 https://lnkd.in/eSNKpwNP #InsurTech #AI #ClaimsManagement #CustomerExperience #HiMarleyJonathan Tushman shared thisWhat if #AI made claims feel more human? In the debut episode of Conversations that Connect, Jonathan Tushman, Hi Marley’s Chief AI Officer and CTO, joins host Stephanie Behnke to explore how AI can eliminate repetitive work, freeing adjusters for the human moments that matter most. Jonathan’s full conversation is now live – be among the first to watch: https://lnkd.in/eSNKpwNP
-
Jonathan Tushman posted thisKeyboards Just Slow Us Down This past weekend, Claude — besides increasing its context window to a million tokens — released speech mode for both Claude Code (hold space to talk in the terminal) and the Mac app (Cmd+G). It's a big deal, and here's why. On January 5th, I had a moment. On Monday I was one version of Jonathan Tushman. By Tuesday I was different. Two things drove that shift: 1. Discovering the power of Claude Code (I know I am not alone here) 2. Starting to talk to my computer (<--- equally impactful) Since that Tuesday, I haven't gone back to typing as my primary input. Every single day, I talk to my machine. If you walk by my office, you'll see me having long-form conversations with my computer. Kicking off work streams. Starting projects. In meetings, Claude is a third participant in the room — we talk to it out loud, ask it to summarize, challenge our thinking. Claude's new speech features are great and getting better. For those who want something higher fidelity right now, I strongly recommend Superwhisper. It uses NVIDIA's open-source Parakeet model, runs entirely locally (nothing goes to the cloud), and it's been absolutely amazing. Here's my ask: as you're going through your own shift in how you work with AI tools like Claude Code, don't stop at changing your code workflow — change your input modality too. Start talking to your machine. Yes, it's going to feel weird. You'll be sitting in your office alone, talking out loud to a computer. Give it a few days. Once you get past that initial awkwardness, you won't go back. This is as big of a deal as Claude Code itself. #AI, #FutureOfWork, #ClaudeCode, #VoiceFirst, #Productivity.
-
Jonathan Tushman shared thisCompliance is a superpower. In the age of AI, the companies that will thrive are the ones who've invested deeply in modern compliance and IT capabilities. This isn't a cost center story. It's a competitive advantage hiding in plain sight. The people manning the walls around your organization — protecting your data, your customers' data, your regulatory standing — their work has never been more important. Every team in your company is about to become an AI-powered innovation engine. The governance infrastructure that channels that energy safely? That's what separates leaders from liabilities. These are tough jobs. They often require saying No. When you seek the happy path -- they are seeking the unhappy paths. The best people in these roles are Protectors, and they deserve to be understood and invested in deeply. I'm lucky to work alongside some of the best — Gabor Tokaji and Dylan McKnight at Hi Marley are two of the greatest. And on the other side of the same coin: your legal team (shout out to Mike Cayer) . They may not man the walls the same way, but they clearly define the boundaries — between you and your customers, you and your partners. THey make you ask the hard questions that you sometimes rather not ask. Their clarity around liability and language is what makes responsible AI deployment possible. If you're a technology leader, I'd encourage you to invest real time understanding what your compliance, IT, and legal partners actually do. The depth of their work is your foundation. https://lnkd.in/eKcAzdSe #AI #Compliance #InsurTech #LeadershipJonathan Tushman shared thisCompliance and IT have traditionally been seen as barriers to speed. In the age of #AI, the opposite is true. Jonathan Tushman, Chief AI & Technology Officer, explains why companies investing in modern #compliance and #IT capabilities are uncovering a competitive advantage hiding in plain sight.
-
Jonathan Tushman liked thisJonathan Tushman liked thisIt used to take a team to get something off the ground. Now, in some cases, it takes one person and a system that can keep up. Polsia is built around that shift. On The Kevin Rose Show, Ben Broca and Kevin Rose talk through what it looks like when a single founder can move from idea to execution without needing to assemble a full team first. Using AI to handle everything from product decisions to go-to-market and ongoing operations. That doesn’t just make things faster. It changes how companies take shape in the earliest days. When the cost of building drops, the spotlight shines on what you're building and why rather than how you'll scale. We’re still early, but this is one of the more interesting unlocks we’re seeing right now.
-
Jonathan Tushman liked thisJonathan Tushman liked thisThat voice in your head isn’t you. Noticing it creates a little distance. And sometimes, that small gap is enough to change your whole relationship to it.
-
Jonathan Tushman liked thisJonathan Tushman liked thisMythica is winding down. I’ve been meaning to write about it for a while, and I want to start with gratitude. Thank you to the Mythica team. Early startups ask too much from people, and I’m deeply grateful for what you gave this one: technically, creatively, emotionally, and personally. Special thanks to Kenny Carvalho. Kenny came in as a consultant and immediately understood what we were trying to build. He helped us find extraordinary technical artists, opened doors, pitched Mythica relentlessly, and stayed with us even when things got hard. Mythica was much better for having him in it. I’m also grateful to advisors, angels, and friends like Dave Taylor, Michael Vorhaus, Nicholas Talarico, Mike Ouye, and Dick Filippini, who were generous with their time, advice, introductions, and belief throughout Mythica’s journey. Thank you to our investors, open-source contributors, artists, technical artists, the Houdini community, Discord members, competition sponsors, customers, and friends. A lot of people gave Mythica their time, trust, code, art, feedback, money, and belief. I’ll always be grateful for that. I’m especially grateful to early customers and partners who took real chances on us: Sprocket Games, Jam & Tea Studios, and Low Drag Labs. Sprocket gave us room to build something beautiful. Jam & Tea brought Retail Mage into the world, and I hope they get to realize their larger AI RPG vision. Minimo is on its way, and I genuinely believe it is going to be special. I’m proud that pieces of Mythica will live on in work like that. The Houdini community also deserves special thanks. They embraced us with open arms, shared knowledge generously, and reminded me how much craft and kindness still exists in technical art. I’m also proud of what we built: open-source infrastructure for games, procedural generation systems, generative asset pipelines, and ways for AI systems to operate real production toolchains rather than just produce text or images. But the deeper thesis was always larger than game assets. AI systems need better ways to understand, generate, and manipulate complex worlds: environments with geometry, materials, constraints, tools, dynamics, and interaction. I still believe that. More than ever. The hard lesson is that being right about the technical direction is not enough. The company, market, financing, and time horizon all have to match the work. In our case, they didn’t quite line up. Mythica was building toward a technical platform that needed time, in a games market going through a very difficult period. Budgets tightened, studios struggled, and many teams had to focus on immediate survival rather than long-term infrastructure. That clarified the real lesson for me: ambition is not enough on its own. The work, the market, the financing, and the moment all have to support each other. Mythica didn’t become the company I hoped it would. But I’m proud of what we built, and grateful to everyone who helped make it real.
-
Jonathan Tushman liked thisJonathan Tushman liked thisIf you’re an engineer or CTO working on agent harnesses and hybrid architectures, this is the event to prioritize during Boston Tech Week. Flybridge is hosting this breakfast panel with Red Hat featuring: - Alexander Amini founder of Liquid AI (one of Boston’s leading AI startups, with $296M raised to date) - Richard Tibbetts, who previously led the agents team at Anthropic - Stephen Watt, bringing the enterprise perspective from Red Hat We’ll go deep on agent harnesses, hybrid architectures, and the real production challenges behind long-running agents. RSVP: https://lnkd.in/emctZxN4 #BOSTechWeek
-
Jonathan Tushman liked thisJonathan Tushman liked thisMalcolm Gladwell said it takes 10,000 hours to become an expert. The big consulting firms are just now starting that clock on AI. We're already past it. That's Eliza. Our engineers weren't introduced to AI on the customer's clock. They've been shipping it for years. The failed agents. The runaway token bills. The human-in-the-loop patterns that actually hold up in production. The 10,000 hours of pattern recognition were already on the meter before our first engagement. And here's what I'm watching now: every project is faster than the last. The trends are real. The Codex movement is real. Token efficiency on the AI SDLC has stopped being a research footnote and become a delivery technique. We're layering expert taste on top of the building. System design. Evals. The call on when to escalate. The output compounds with every retrospective. Per project. Per meeting. More value, in less time, at lower cost. This is the part the legacy consulting playbook is going to miss. Big firms have their own 10,000 hours. But they were logged in tax, audit, ERP rollouts. Anywhere but here. Now they're trying to flip the script and reroute that experience into AI services. You can hire AI talent. You can stand up a practice. What you can't do is retroactively be born for this moment. One team. One focus. Ten thousand hours already in the bank, compounding daily. That's the unfair advantage our customers get.
-
Jonathan Tushman liked thisJonathan Tushman liked thisGenuine question: who out there in my network is also fine-tuning, quantizing and running models for practical workloads? Let's talk!
Projects
-
babar
-
Babar is an intelligent codebase analyzer designed to bridge the gap between your existing codebase and modern AI-powered development workflows. By recursively analyzing your project's directory structure and creating smart summaries, Babar helps both developers and AI tools better understand and work with your codebase.
-
bouncer
-
See project**bouncer** is an authorization library that restricts user access to resources. All the permissions are defined in a **single location**.
* Decoupled from authentication
* Decoupled from how you implement roles
* Database independent -
memory_utils
-
See projectMemory Issues happen to the best of us. `memory_utils` will give you simple tools to quickly isolate the culprit, and ideally, warn you before you run into issues.
From my experience, there is no silver-bullet in dealing with memory issues. You just have to roll up your sleeve and get dirty with print statements. In our team's recent fight with a memory issue, we created memory_utils and we wanted to share.
`memory_utils` deals primarily with RSS memory (Resident Set Size). The…Memory Issues happen to the best of us. `memory_utils` will give you simple tools to quickly isolate the culprit, and ideally, warn you before you run into issues.
From my experience, there is no silver-bullet in dealing with memory issues. You just have to roll up your sleeve and get dirty with print statements. In our team's recent fight with a memory issue, we created memory_utils and we wanted to share.
`memory_utils` deals primarily with RSS memory (Resident Set Size). The most important memory concept to understand when dealing with memory constrained systems: RSS, the resident set size, is the portion of a process's memory that is held in RAM. The rest of the memory exists in the swap of the file system.
-
monarch
-
See project`monarch` is a mongo utility belt that helps developers and admins deal with common admin use-cases.
The main use-case that this library was built for was _migrations_ but it does a bunch or other useful things like makes it easy to backup, restore, and copy environments between one another.
It has been very helpful for our teams -- and hopefully others will find it useful as well. -
state_machine
-
See projectThere are two types of developers in this world: those who love state machines and those who will eventually.
I fall in the first camp. I think it is really important to have a declarative way to define the states of an object. That’s why I developed `state_machine`.
Other similar profiles
Explore more posts
-
Sam Somashekar
Health-Insurtech: The Next SaaS Frontier 🩺 Insurance operations are finally getting their SaaS moment. This week, SehaTech, a Cairo-based health-insurtech startup, raised $1.1M in seed funding to automate claims, policy administration, and provider management in frontier markets. That might sound like a niche round, but it represents a much bigger signal: The next SaaS frontier in health-tech isn’t new apps or devices - it’s automation of the back office that powers healthcare finance. Here’s why this matters 👇 🔹 1. The biggest inefficiencies are invisible. Across emerging markets, insurers and TPAs still run claims processing on spreadsheets, emails, and manual validation. That’s a billion-dollar friction zone ripe for automation, and SaaS is finally catching up. 🔹 2. The “fintech-ization” of health-tech is accelerating. The same playbook that transformed banking (digital rails, transaction validation, and automation) is now being applied to healthcare payments. Insurtech startups like SehaTech are using AI and APIs to link providers, payors, and patients into real-time claims ecosystems. 🔹 3. SaaS for regulated operations = long-term moat. When your platform becomes the compliance layer for policy management or claims adjudication, you don’t just have users; you have infrastructure lock-in. That’s why even small insurtech automation rounds deserve outsized attention from investors. 🔹 4. The cross-industry signal. Fintech solved the “money movement” layer. Health-tech SaaS is now solving the “data and payment integrity” layer. And that convergence - finance logic + healthcare regulation - is where the next enterprise SaaS platforms will emerge. The opportunity is massive: To build the software that makes healthcare finance work faster, cheaper, and more transparent. 💬 What do you think: will health-insurtech be the next wave of B2B SaaS growth in emerging markets, or does regulation still slow it down? Let us know! #HealthTech #Insurtech #SaaS #DigitalHealth #Fintech #EmergingMarkets #Automation #AI #ProductLeadership #VentureCapital #HealthcareInnovation #ProductManagement
-
Saran S.
I sat down with Frontegg CEO Sagi on scaling CIAM for Insight Health. A few highlights from our convo, • Day-one enterprise readiness (HIPAA, multi-org, customizable access policies) • Why “build your own auth” slows growth (SSO/RBAC/SCIM, audit logs) • Delegated admin so partners self-manage without pinging engineering • CIAM as a GTM enabler for AI agents and clinician workflows Replay + transcript: https://lnkd.in/grRed-C5
1 Comment -
Azam Khan
NEW: NY’s OCM has decided to transition to Metrc for their seed-to-sale tracking system. This comes after the partnership between Metrc and BioTrack, BioTrack of course was the tech company initially given the seed-to-sale contract in New York. Things to keep in mind as per https://lnkd.in/g_gunAbG: • Licensees who were sending data via the BioTrack NY STS API can now stop. • All licensees are still required to maintain an electronic, real-time inventory tracking system. • Metrc UIDs will cost $0.10, and any BioTrack tags previously purchased will be credited. • Third-party integrators will not be charged to transition from BioTrack’s API to Metrc’s API. • Labs will send testing data to Metrc under the new system. • Until Metrc goes live, transfers must use paper manifests, and licensees should keep historical data in their own systems. • OCM and Metrc will provide updates on API access, sandbox testing, training, and tags in the coming months. • Operators are advised to continue to submit inventory/sales reports to the Office via the portal. • They are aiming to help operators go live with Metrc in early 2026 Distru is a validated third party software integrator with BioTrack in NY, and we are committed to working with Metrc as OCM figures out what next steps are in the coming months. #NYCannabis #Metrc #SeedToSale #Compliance #CannabisTech #Distru
3 Comments -
SC Moatti
I’m excited to share that Angle Health, founded by former Palantir Technologies execs Tylon (Ty) W. and Anirban Gangopadhyay, has announced an oversubscribed Series B bringing total funding to nearly $200M. I’m proud to be an early investor and to have led this competitive round with the amazing team at Portage. Why we keep leaning in: - 26x revenue growth since the Series A - Serving 3,000+ employers across 44 states - 80%+ renewal rates Angle Health is rebuilding the healthcare benefits infrastructure for SMBs, who employ nearly half of America’s workforce but have historically been shut out of enterprise-grade coverage. Their AI-native platform helps predict risk and deliver better care at a sustainable cost. This is exactly the kind of product-driven, AI-native company we’re proud to support at Mighty Capital.
-
Insurtech Universe
Insly’s launch of NORA signals something important about where insurance AI is heading: From automation to orchestration. Rather than positioning AI as a standalone feature, NORA is built as an AI layer — designed to systematically remove operational bottlenecks across underwriting and policy workflows. This reflects a maturing mindset in insurtech: • Less hype. • More embedded intelligence. • Real workflow impact. The insurance industry is full of friction points — fragmented systems, manual bordereaux processes, regulatory reporting burdens. Platforms that eliminate those inefficiencies at scale will create structural advantage. AI in insurance isn’t about replacing expertise. It’s about accelerating it. 🔗 Full article: https://lnkd.in/gXuvjqGZ #InsurTech #AI #InsuranceOperations #MGAs #DigitalInsurance #FutureOfInsurance
1 Comment -
Pranav Arora
We didn't plan to become an enrollment solution. Last month, 3 Medicare brokers filled out our contact form. Different states. Same question: "Can your voice AI handle enrollment outreach?" We built Rivvi for primary care - an automated voice AI that calls patients about medication adherence, chronic care management, and other patient communication. Qualifies them, books appointments, and handles follow-ups at scale. Then the navigator budget cuts hit, and brokers started drowning in manual outreach during AEP. We already had: - Voice AI is calling thousands of patients daily for CCM programs - Dynamic workflows that qualify and filter by criteria - Appointment booking and callback management - Inbound and outbound patient engagement campaigns - Systems that handle seniors asking the same question 3 times Enrollment outreach is just another workflow variation. Same tech, different qualifying questions. Configured the first one in a day. Not built. Configured. We can run multiple broker campaigns simultaneously. Setup takes 2-3 days - mostly intake and workflow customization. The technical challenge wasn't adaptation. It was recognizing the problem. The same AI that keeps seniors on their medications can handle the mass outreach brokers are drowning in during AEP - so they can focus on actually helping people who need enrollment guidance instead of working for free, making manual calls. If you're dealing with this, comment ENROLL. System's already running. #MedicareAdvantage #HealthcareAI #InsuranceBrokers
-
Frank Pepper
This is the auto-healing system inside VantageOS. Most platforms depend on people to notice when something breaks. This one is built to notice, understand, and respond on its own. Behind the interface is an internal agentic layer continuously monitoring system state, data integrity, workflow execution, and operational signals. When something drifts, degrades, or fails, it does not just raise a flag. It diagnoses the issue, traces dependencies, and initiates corrective action to restore alignment. Quietly. Continuously. Without waiting. Because in complex environments, small inconsistencies do not stay small. They cascade. Adam and I built this with a simple principle: a serious system should not just run, it should maintain itself. That means detecting anomalies early. Resolving issues before they compound. And keeping the platform stable, reliable, and aligned even as conditions change. This is what operational resilience looks like when it is built into the system itself. #VantageOS #AI #AgenticSystems #Automation #OperationalIntelligence #Resilience #Infrastructure #SystemsThinking https://lnkd.in/giTUvp3p
-
Prateek Joshi
AI agents need an insurance product. And Artificial Intelligence Underwriting Company (AIUC) is aiming to build that. Founded by Anthropic's first product / GTM hire. They're coming out of stealth with a $15M seed led by Nat Friedman. To underwrite AI agents, there are 3 points to think about: (1) need to model failure modes and price them dynamically with live telemetry (2) need to enforce continuous certification, audit‑grade observability, and tiered coverage (tied to autonomy level) (3) need to align with changing AI regulations and layer reinsurance to manage systemic risks (and satisfy capital requirements)
-
Sas Ponnapalli
A recent Becker's Healthcare article talked about AI becoming a real growth driver for health systems. 🔗 bit.ly/3NLQAKr What stood out is where that growth is actually happening, in everyday workflows like scheduling, intake, and coordination. That’s where we’re focused at Beam, and why we just introduced AI Caller, a voice-based AI Agent that handles scheduling from start to finish, verifies insurance, and captures clean intake from the very first interaction. For teams, it means fewer interruptions and less manual back and forth. For patients, it’s faster, simpler, and feels more natural with voice options that sound like a real person. Nothing complicated. Just a better way to keep things moving. #BeamHealth #AIinHealthcare #PatientAccess #HealthcareOperations
1 Comment -
TX Zhuo
For years, AI in insurance meant automating the back office: faster quotes, smoother workflows, fewer manual tasks. That was a great start, but speed alone won’t win in an increasingly competitive market. What excites me now is the shift toward AI that actually drives revenue. Tools like Outmarket AI, CoverForce, and Liberate are helping brokers prioritize high-conversion submissions, align with carrier appetite, and re-engage dormant leads. It’s not just about efficiency anymore. It’s about enabling smarter growth. At Fika Ventures, we’re excited by platforms that combine automation with intelligence. The winners in this space won’t just move faster. They’ll move with precision. Read more of my thoughts in the blog below 👇 https://lnkd.in/gNZ255GQ
8 Comments
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content