Upstage

Is your LLM struggling to "read" complex documents? Most models fail because they only see text, not structure. Our Visual Document Extraction uses a specialized "eyes and brain" architecture: • The Eyes (OCR): Captures visual components like tables and layouts • The Brain (LLM): Interprets the data once it's in a language it understands Check it out in action: https://lnkd.in/g96uiUKj  

Upstage has signed a strategic Memorandum of Understanding (MOU) with HP Korea, advancing AI agent technology for the enterprise. Under this partnership, Upstage and HP Korea will combine HP's AI workstations with Upstage's proprietary Solar LLM to develop enterprise-grade AI agent solutions, purpose-built to transform how organizations work and compete in an AI-driven world. Learn more: https://lnkd.in/g39TkFrQ

Broker packages are messy by nature. ACORD applications. Loss runs. Different formats. Missing fields. Details that matter, buried across multiple documents. In this interactive demo, Kasey Roh shows how Upstage Studio helps turn that intake work into a repeatable automated workflow. You can see the process from start to finish: → Upload the submission documents → Describe what needs to be pulled → Review the auto-generated schema → Trace extracted fields back to the source → Connect the output to external systems And that is not even the full flow. You will also see how teams can create an agent, review extraction results, edit fields, mark information as correct, and tune the workflow for future runs. Studio is not just extracting information from a document. It is giving teams a way to build, review, refine, and deploy document workflows that fit the way the work actually happens. That is what makes Studio different. See the workflow in action → https://lnkd.in/gZgRpVXr #UpstageStudio #UpstageAI #DocumentAI #UnderwritingAutomation #InsuranceAI #AIWorkflowAutomation

Upstage CEO Sung Kim took the keynote stage at Amazon Web Services (AWS) Seoul Summit 2026. AI agents are everywhere, but most enterprise teams are still doing the work by hand. To actually fit into business operations, agents need reproducible outputs, audit trails, step-level error correction, and manager approval. That's the problem Upstage is tackling with Upstage Studio, our agent builder. The approach is already paying off. In one deployment, document processing that used to take over 20 minutes is now done in under 5, a case Sung walked through on stage. The numbers back it up too. Upstage Document Intelligence is now ranked #2 most subscribed product in AWS Marketplace across AI Agents & Tools category. We'll be at COEX 3F C Hall, Booth B-30 all two days — stop by for a live Studio demo and chat with our team. Build your first document agent in minutes at Upstage Studio. https://lnkd.in/gVYbEfPi #AWSSeoulSummit #UpstageAI #EnterpriseAI #AIAgent #Upstagestudio

“End-to-end AI” sounds like the clean answer to fragmented workflows. One system, one flow, everything handled in a neat, linear process. The problem is that insurance doesn’t actually work that way. Workflows vary by line of business, decisions depend on context, and exceptions are part of the process, not edge cases. What looks seamless in a demo starts to break down in real operations, where variation is constant and judgment still matters. That’s when teams begin working around the system instead of through it: — Manual steps reappear — Workarounds pile up — The “simple” solution becomes another layer of rigidity. The bigger risk isn’t choosing the wrong tool. It’s getting locked into one that can’t evolve as the workflow or the technology changes. Article 3 in our series breaks down why end-to-end AI falls short in real insurance workflows and what that means for teams trying to scale it. #InsuranceAI #Underwriting #Claims #Insurtech

Insurance Innovators is a good read on where this industry's head is at. The companies gaining ground are not talking about AI in the abstract anymore. They are talking about specific workflows, specific problems, and which partners have actually delivered. That is a different conversation, and a more important one.

Most insurance teams don’t have an AI problem. They have a fragmentation problem. It starts small: — One tool for underwriting — Another for claims — A script here, a workflow there Individually, they work. But together, they don’t. Over time, teams end up managing dozens of disconnected workflows with no shared structure, no consistent outputs, and no clear way to trace decisions. That’s where the real risk shows up. Article 2 in our series breaks down how fragmentation builds and why it becomes a system-level issue. #InsuranceAI #Underwriting #Claims #Insurtech

Upstage is at Government & Public Week 2026 in Tokyo this week, exhibiting alongside HP Japan with Upstage Enterprise Workspace, our enterprise AI Agent platform. Upstage Enterprise Workspace is a fully self-contained, local AI solution built for the security requirements of Japan's public sector, one of our strategic focus areas in the market. If you're at Tokyo Big Sight through Friday, stop by HP Japan's booth (West Halls 1-2) — Hiroyuki (Hugh) Matsushita from our Japan team will be there to welcome you. #Upstage #SolarLLM #PublicSector #EnterpriseAI #TokyoBigSight

Trying to do everything at once is what slows most teams down. The ones making progress are focused, deliberate, and building from a single problem outward.

We’re asking the wrong questions about automation in insurance. Most conversations still start with the technology. What tool to use, what vendor to evaluate, what features to compare. The teams actually making progress tend to start somewhere else. They focus first on the problem they’re trying to solve and how work actually flows across their organization. Only then do they look at where automation fits. When that order is reversed, things get messy. Solutions are introduced before the problem is clearly defined, which creates complexity early and makes it harder to scale later. It’s how systems can look effective in a demo, but struggle once they’re applied to real workflows. It also explains why “end-to-end” automation sounds appealing, even though most insurance workflows don’t operate that way in practice. Insurance workflows don’t follow a clean, linear path. So when the conversation starts with “How do we apply this?”, it skips over the more important question: what part of this process actually needs to change? That’s the gap we’re trying to close. If what you’re seeing in practice doesn’t match what’s being sold, you’re not imagining it. Explore the full piece → https://lnkd.in/gRp65qNH