> ## Content Index > Fetch the complete content index at: https://blog.durst-group.com/llms.txt > Use this file to discover other available public pages before exploring further. # Building, in Beta. - URL: https://blog.durst-group.com/building-in-beta/ - Published: 2026-10-01T11:16:14.000Z - Updated: 2026-10-01T11:16:14.000Z - Description: Ink that expires on the shelf. Presses that stop without a reason. Four new Kyveris features from PRINTING United, built with the people who print. - Author: Michael Deflorian - Tags: #kyveris, #chapters 01Mindset ## One word: building. We keep saying *building*. On purpose. [Kyveris](https://blog.durst-group.com/origin/) is how Durst connects machines, software, data and AI in digital print production. *Building* is how Kyveris becomes reality. Building is messy. Building is someone saying “this doesn’t work yet,” and everyone learning in instead of looking away. We like it that way. *Beta* is building’s more social sibling. Beta means we don’t do this alone in our labs. Everything is built together with our beta customers, on their print floors, with their equipment, their shifts and their real feedback and experience. They see it early. They tell us what’s broken, what brings value and what doesn’t. We make it better and come back. We keep ourselves honest with three questions. (I) Does it work? (II) Can someone who didn’t build it use it? (III) Are the results good enough to act on? Let’s look at how we do it, and what came out of it. 02Principle ## Problem first. Always. We keep one thing in mind. Before we build anything, we identify a problem. A real one, on a real print floor, ideally one that has annoyed a print shop on more than one Monday. Then our passionate digital print pioneers start building. 03Kyveris ## A city, not a building. A building gets finished. There’s a day you hand over the keys. A city never has that day. New streets connect to the ones already there, and every new district makes the old ones more useful. Kyveris works like a city. It isn’t a finished product. It’s Durst’s path into the future of Digital Print Production. At PRINTING United Expo in Las Vegas, we opened four new streets from three districts/domains. They connect to what’s already running. | 01 | Machine learningInk Forecasting | | ------------------------------------------------------------------------------------------ | ---------------------------------------------------------------- | | 02 | Connected dataDurst Insight | | 03 | Conversational AIDurst Service Chatbot · Durst Analytics Chatbot | | Next | TwinsMachine and process, simulated | | Kyveris at PRINTING United Expo 2026: three domains, four new features, one next district. | | None of them is finished. All of them are running in Beta. More will follow. And always enjoy the ride! 04Machine learning ## Domain one: Machine learning. Here’s the one-sentence version. Machine learning is software that studies a machine’s or process past so closely that it starts to see the future coming. No fixed rules. No “if this, then that.” Just patterns, always grounded in data, and a model that gets sharper the more it sees. A press produces data every second, but data on its own only looks backward. Machine learning turns it around. Ink Forecasting is the first Kyveris feature built on this understanding. Data-Architecture was the pain for fleet wide roll-out. It won’t be the last. ### Ink Forecasting. Beta statusIt works, and it gets better with every iteration. Ink is patient. It sits in the machine or on a shelf and waits to be jetted. And while it waits, it ages. Quietly, with no warning light. Stock is a balancing act. Too little ink stops production or forces an expensive express delivery. Too much, and it expires: money lost, ink thrown away. Left in the machine, expired ink can damage the printheads, and during the warranty period it voids the warranty for that channel. Now multiply that by a fleet. Each press has its own ink stock, its own jobs, its own consumption. One press burns through cyan. The press next to it has been holding the same light magenta for weeks. Think of a restaurant kitchen. A good chef checks the fridge before writing the menu. Whatever expires first gets cooked first. The menu follows the stock, so nothing goes to waste. Ink Forecasting does the same for a fleet of presses. A machine learning model from ADA’s AI Squad learns from past print jobs how much of each ink each press consumes. Once a day, it forecasts consumption for the weeks ahead, and whether the ink on a press will be used up before its expiry date. That turns reaction into planning. Jobs that use a lot of a soon-expiring ink run first, on the press that holds it. An ink change or service visit lands in a quiet production window, not under time pressure. More ink turns into prints instead of waste, and the whole fleet is put to better use. | Ink Forecasting at a glance | | | ---------------------------------------- | ------------------------------------------------------------- | | 01 | LearnConsumption per ink and per press, from past print jobs. | | 02 | ForecastOnce a day, for the weeks ahead. | | 03 | CompareForecast against stock and expiry date. | | 04 | ActJobs, ink changes and service visits, planned in time. | | The forecast appears in Durst Analytics. | | The same forecast is the basis for operator notifications. And the model is deliberately simple: easy to trust, and ready to forecast more than ink. Under the hood, press data, the forecasting model and Durst Analytics are connected in one pipeline, closed end to end in development and moving into production next. Every model that follows will use it, and the model gets better the more data it collects. 0:00 /0:08 1× Ink Forecasting: ink demand, plannable before stock or expiry bites. 05Connected data ## Domain two: Connected data. Monday morning, production meeting, somewhere in the US. Someone asks: “Why did press two sit idle twice on Thursday, three hours in total?” Silence. Maybe it was material. Maybe a file was late. Maybe someone remembers, but that someone is on the night shift. So the discussion runs on gut feeling, and nobody can say which bottleneck to attack first. A print shop runs on presses, finishing, workflow software and **people**. Each keeps its own diary, and nobody reads them all. We think every one of those diaries belongs on the same table: open, transparent and owned by the customer. ### Durst Insight. Beta statusEvery source, one view. Durst Insight connects sources that used to be closed systems. Machine data and operator data now come together in one view. The operator’s part is built into daily work. When a press sits in setup or idle longer than a configurable threshold (5, 7.5 or 10 minutes), a block screen appears. The operator picks a downtime reason to unlock it. The reason is captured the moment it happens, not remembered at the end of the shift. It lives in the system, classified. | Durst Insight at a glance | | | --------------------------------------------------- | ---------------------------------------------------------------- | | 01 | DetectA press sits in setup or idle beyond 5, 7.5 or 10 minutes. | | 02 | AskA block screen appears on the press. | | 03 | CaptureThe operator selects the reason to continue. | | 04 | AnalyzeLabeled causes become trends and visible bottlenecks. | | Every stop gets a reason, at the moment it happens. | | This is what makes the data usable. Think of a sleep tracker. One restless night tells you nothing. Three weeks of nights tell you it’s the late coffee. Insight works the same way. Idle time, and how it trends over days and weeks. Causes that keep coming back. The real bottleneck, whether it hides in the press, the material or the handoff just before it. Production gets better because decisions rest on patterns, not guesses. 0:00 /0:04 1× Durst Insight: unexplained downtime, turned into labeled causes. Next Monday, nobody has to go quiet. The answer is on the screen. 06Conversational AI ## Domain three: Conversational AI. Think back to the first car you drove. A light starts blinking on the dashboard. It’s orange, and it’s shaped like something you’ve never seen before. You pull over, open the glovebox and dig out a manual the thickness of a phone book. You flip. You squint. You guess. Now think of a dream world today. You just ask: “What does that light mean?” The car tells you in plain language what it is, how urgent it is and what to do next. Same car. Same light. A completely different experience. That’s conversational AI. You ask the way you’d ask a colleague, and you get a precise answer to your specific problem. No more knowing which menu or which page to look in. Now picture a print floor at 2 a.m. A press shows an error. The manual has hundreds of pages. That’s exactly where print needs it too. What matters most is where the answer comes from. Our conversational AI works on Durst’s own knowledge: our documentation, our service history and data from years of global experience. Every answer is grounded in those sources, and it’s about your machine. We’re starting with two features in this direction. ### Durst Service Chatbot. Beta statusIn use inside Durst. First customers are next. When a machine goes down, the knowledge to fix it is spread across manuals, past tickets and the heads of the few technicians who have seen that fault before. Meanwhile, the customer is standing at a stopped press, often with a problem they could have solved in minutes. The Service Chatbot puts the most experienced technician into every service call. It has read everything: manuals, service documentation and years of service tickets. The tickets are the secret ingredient. Documentation tells you how a machine *should* behave. Tickets tell you how it *actually* behaved, and what finally fixed it. The chatbot knows both. It lives in the service portal, now connected to MyDurst, and it starts with our own service technicians. In the coming weeks, the first two customers with self-service training start testing it, followed by a rollout to our customer engineers. The goal: customers solve common issues themselves instead of waiting for a visit. 0:00 /0:08 1× The Durst Service Chatbot in the service portal. ### Durst Analytics Chatbot. Beta statusThe conversation works. The data comes next. Durst Analytics collects a lot of production data. But reading it is a skill of its own, and for everyone who isn’t an analyst, the value stays locked in the numbers. The Analytics Chatbot turns dashboards into answers. You just ask. It’s built as one shared foundation for Durst’s software products: one interface component, one backend. The conversation feels the same everywhere, and nobody has to build it twice. Today, it knows the product documentation, so it explains what a visualization shows and what it’s meant for. It doesn’t read your production data yet. That part is being built right now, deliberately in that order: first the conversation has to feel right, then it gets the data. 0:00 /0:05 1× AI-generated illustration. 07Outlook ## A glimpse out of our labs: The twin. Imagine a press that exists twice. One stands on the floor. The other lives in software and behaves exactly like the first. Today, the only way to know how a job will run is to run it. Every test costs material, machine time and nerves, with no guaranteed outcome. With a twin of the machine and the process, you try first. A new substrate, a tricky job, a bold setting: all of it goes into the model before it touches the press. Once the twin is accurate, it connects back into production, so what the simulation learns flows straight onto the floor. It’s the next district of Kyveris, where machine learning, connected data and conversational AI meet. We can’t wait to show you. Stay tuned. > Kyveris isn’t something we hand over. It’s something we build side by side with our customers, for the industry we shape together. **Michael Deflorian* — Vice President Kyveris* 08Join ## Build it with us. The city keeps growing. These products get better every time someone on a real print floor uses them and tells us the truth. If you want to join this group, join the beta program. [Join the beta program →](mailto:Tomio.Kiyose@durst-group.com?subject=Kyveris%20Beta%20Program) *— Michael Deflorian, October 2026*