Upstage

🎙️ New Episode | The Leadership In Insurance Podcast I had a brilliant conversation with Kasey Roh, Head of US at Upstage AI, and if you work in underwriting, claims, or insurance technology, this one's a must-listen. Kasey's background is unlike most people I speak to in this industry. Having worked through periods of hyper-growth at Meta and Tesla, she saw first-hand what happens when organisations try to automate before they truly understand their own processes, and the costly lessons that follow. It's a mindset she's brought directly into how Upstage helps insurance carriers adopt AI the right way. We got into some really important territory: 🔹 Why mapping your workflows before automating is still the most overlooked step in AI adoption 🔹 How Upstage went from powering 60–70% of claims processing in Korea to taking on the US enterprise market 🔹 What OCR and LLM technology is actually doing for underwriting and claims — and why legacy systems are struggling to keep up 🔹 Whether it's existing carriers or new AI-native players who will ultimately reshape the industry Kasey's take on where insurance AI is heading is candid and grounded: the proof-of-concept phase is ending, and the organisations without proper governance and human oversight are going to find the scale-up very difficult. 🎧 Links in the comments 👇 I'd love to hear your thoughts.

【Upstage - SusHi Tech登壇 ＆ Studioを日本市場ローンチ発表】 Upstageは、本日から開催されるSusHi Tech Tokyo 2026 on April 27-29 登壇に先立ち、AIエージェント基盤「Upstage Studio」を発表しました。請求書・契約書・保険金請求書 — ドキュメント処理が人手作業で課題になっている業務を、ノーコードで Agenticなワークフローに置き換えるプラットフォームです。 AI導入がPoCで止まる原因は、モデル性能ではなく「現場のドキュメントをAIが使える形にできるか」ー Studioは、そこに正面から取り組むプロダクトです。イベント会場は本日夕方に高市総理も登壇されますが、私も明日28日のAtrium Stageで他のAIユニコーン企業の皆様と並び、Pitchの機会を頂いてます。 今回、会場ブースでは、Upstage初の試みとしてお米（コシヒカリ）をSWAGとして配っています！是非、会場でお会いできるのをお待ちしております!! 📍 Atrium Stage｜4月28日(火) 12:50〜 📍 West Hall 1, C539 Launched today at SusHi Tech Tokyo 2026: Upstage Studio — a no-code, AI-powered platform designed to build, deploy, and manage #agentic workflows for invoices, contracts, claims, and other document-heavy processes. If your team is still manually processing invoices, contracts, or claims — come find us at SusHi Tech Tokyo this week! We're showcasing the Upstage Studio in Japan at SusHi Tech. Today, Prime Minister Takaichi takes the Stage at SusHi Tech Tokyo. Tomorrow (April 28), I'll have the honor of standing on the Atrium stage — alongside other AI unicorn leaders — to share Upstage's story and what we're building for Japan. Come get the most useful swag (Rice) you'll get at any tech conference this year! 📍 Atrium Stage · April 28 · 12:50 📍 Booth: West Hall 1, C539 Press Release: https://lnkd.in/gS2vurYv

Every insurance workflow I've seen starts in the same place — documents. Submissions arrive as PDFs, emails, spreadsheets, attachments stapled together in ways nobody planned for. Some of that data makes it into systems. Most of it doesn't. I've talked to teams where 90% or more of the information in a submission never gets structured — it just stays trapped in the file. And then everything downstream gets built on top of that incomplete foundation. Models, workflows, business logic, decisioning — all running on inputs that were never fully captured in the first place. That's the problem we decided to solve. We sit at the very front of the process — turning unstructured documents into structured, reliable data at scale. Because once that foundation is solid, everything that comes after it actually has a chance of working. We'd love to hear about your document chaos: https://lnkd.in/gxAiXNsJ

General purpose AI can do a lot of things well. But domain specific tasks, like underwriting, isn’t a general purpose job. When teams rely on broad LLMs, they often end up forcing complex workflows into a small prompt window, hoping the model understands the context, the documents, and the decision criteria. Sometimes it works. But it also puts a lot of burden on the team to shape the input just right. Purpose built systems change that dynamic. When a model is designed around document heavy workflows and a specific domain, much of the context is already built in. That means less prompt engineering and more time focused on the decisions that actually matter. Great perspective from Amwins Benefits Vice President of Digital Strategy Steven Beauchem! https://lnkd.in/e5qCuVVb #InsuranceAI #Underwriting #InsurTechStrategy #InsuranceOperations

Wrapping up a few quick but packed days at the Auto Insurance Report National Conference in Dana Point — a beach run at sunset, and some of the brightest minds in auto insurance all in one room. You really can't beat the combination. A few sessions that stuck with me: - Bill (LexisNexis Risk Solutions) on how introspection and a willingness to keep learning is what lets a business continue to innovate. Easy to say, hard to actually do — and a great reminder that humility is a competitive advantage. - Ryan and Michael (Connected Analytic Services and Toyota Insurance) on how connected vehicle data can create genuinely meaningful experience differences for insurance customers. The gap between "we have the data" and "we've done something useful with it" is where the real work lives, and Toyota is clearly doing that work. - Brad Nail (Waymo) on how autonomous vehicles are already improving road safety today. As a consumer, I walked out more excited than ever for the day full autonomy is just part of life. Big thanks to Pat and Brian for putting this together year after year — the content, the crowd, and the care that goes into it really show. And a shoutout to my partner in crime on the road, Brian! #AIRI2026 #AutoInsurance #InsurTech

Four days at the Motor Carrier Insurance Education Foundation (MCIEF) Forum in Orlando with my partner in crime Brian — and I'm leaving with a real appreciation for this community. MCEIF brings together a focused group of people who live and breathe commercial auto and trucking insurance. The conversations here weren't about AI hype or what's trending. They were about how painful loss runs still are, what it actually takes to get clean data out of aging systems, and the day-to-day grind of keeping commercial auto profitable. That kind of specificity is where the best conference conversations happen. Two things I won't forget: - The afternoon ice cream station. I don't know who made that call, but every conference in America needs to steal this idea. There's something about handing someone a bowl of ice cream that makes the next conversation twice as good. - And getting serenaded by John with a full band. I did not have that on my conference bingo card. The man can sing. Big thanks to the MCEIF team for putting together a genuinely warm, well-run event. Heading home to catch the kids before bedtime. On to the next one.