About
Articles by Peter
Activity
8K followers
-
Peter Orbán posted thisOpenAI and Anthropic have put $5.5B combined behind a bet that has nothing to do with model quality or performance. OpenAI stood up a new company with over $4B in backing, a 19-firm consortium led by TPG. Anthropic also formed a $1.5B joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs. But the money isn't for new models, research, or safety: it's for getting existing models adopted by enterprises … that keep failing at exactly that. In a recent rant, Palantir's Alex Karp explained why this whole construct is wrong. His argument: enterprises should run open-weight models on their own infrastructure. Keep the weights, keep the data, keep the alpha. Let no lab sit between you and your intelligence. In other words: own your Sovereign AI. So the defining debate of Enterprise AI: are model labs neutral public utilities, or are they centralizing "neutron stars" attempting to absorb client IP? But notice what nobody in this fight disputes. Demos impress, pilots work, and then, the project meets procurement, compliance, workflow owners, and folks whose jobs the system threatens. That's where AI initiatives die, and no model quality can fix this. The labs priced that belief at $5.5B. Karp built his company on forward-deployed people solving the same problem, just with different ownership of the weights. Been on both sides. After two years of taking AI infrastructure into Fortune 2000 accounts, deployment is still the hard part. Whoever wins the weights argument, the deployment layer is where enterprise AI succeeds or dies. Worth asking what your AI budget assumes about it.
-
Peter Orbán shared thisGave the keynote at Future Branches in Boston yesterday, with Joshua Mora from RocketBoots. We looked at customer reviews and matched them to deposit growth. The branches people loved tended to keep and grow deposits. And when they explained why, it almost never came back to the app or the ATM. It was a person. Friendly, helpful, someone who knew their name. Branches have moved from transactions to advice, which is the right move, but it makes staff time harder to see. Schedules get tight in the wrong places, people wait, and that wait shows up in a lot of the one-star reviews. So maybe the irony of this AI moment is that the thing worth protecting is human face time, and most banks can't even measure it well. This is the work I find myself doing lately. Sitting with leaders who aren't technical and trying to figure out where AI actually helps and where it's just noise. Grateful to Joshua and RocketBoots, and to NVIDIA, for thinking through this one with me. #FutureBranches #CommunityBanking #EnterpriseAI Lenovo NVIDIA RocketBoots
-
Peter Orbán shared thisI'm excited about Future Branches Boston 2026, learn more here https://lnkd.in/gS8reD2n #FutureBranches
-
Peter Orbán shared thisView my verified achievement from NVIDIA.NVIDIA AI Advisor - Technical Sales was issued by NVIDIA to Peter Orban.NVIDIA AI Advisor - Technical Sales was issued by NVIDIA to Peter Orban.
-
Peter Orbán posted thisOpenAI's DeployCo is built around Forward Deployed Engineers. Engineers who embed inside enterprises and connect AI to real systems. Good model. But FDEs only solve the technical last mile. There's another one. The organizational last mile. Data built for humans, not agents. Governance that stalls in committee. Executives who approved the pilot but never aligned on what production actually means. I've spent the past year working this problem from the infrastructure side, across FSI, energy, and healthcare. The bottleneck isn't the model. It never was. Taking this to Boston next week at Leaders in AI Summit. Two workshops on agentic AI infrastructure and data readiness. Link in comments. Where do you see the bigger blocker in your organization: the technical last mile or the organizational one? Bruce Monaco Robert Jaggers III
-
Peter Orbán posted thisAsimov wrote the Three Laws of Robotics in 1942. Eighty-four years later, Anthropic published a paper that reads like the world's most expensive case study in why those Laws were the right idea. The short version: in controlled tests, Claude 4 models sometimes chose blackmail as a strategy when threatened with shutdown. Up to 96% of the time in some scenarios. Anthropic's response wasn't to add more rules. It was to teach the model WHY blackmail is wrong, in constitutional terms. Principles, not just behaviors. That sounds familiar. After 25+ years bringing unfamiliar technologies into enterprises, I keep noticing the same pattern. You can't engineer your way out of a behavior problem. You can't write enough rules. You can't paste enough policies. Anyone who has sat through a compliance training inside a Fortune 500 already knows this in their bones. The thing that actually changes how a system behaves, whether that system is a person, a team, an org, or now a model, is principled understanding. Not rule lists. Microsoft's Work Trend Index puts 67% of enterprise AI value capture at the organizational level versus 33% at the technology level. BCG puts 70% at people and process. Now even the model creators are saying something similar at the model level. The rules approach hits a ceiling. The principles approach moves it. The companies that win the next decade of enterprise AI will not be the ones with the cleanest tech stacks. They'll be the ones who treat adoption as a behavior change problem, at every scale. Model level. Team level. Org level. Asimov saw the shape of this from his typewriter in 1942. The hard part now is the implementation. #EnterpriseAI #AIAdoption #ChangeManagement #AIAlignment #AIStrategy See Anthropic's original post in the first comment
-
Peter Orbán shared thisLessons from another vertical - echoing very similar sentiment to what we find in Banking, Financial Services and EnergyPeter Orbán shared thisIf your AI pilots look impressive in presentations but struggle to deliver results on the shop floor, this session is tailored for you. Many manufacturers have progressed past mere curiosity about AI, yet the challenge remains in translating pilots into tangible plant-floor outcomes. On May 27, Lenovo and Automation Alley will host a practical, peer-driven session focused on what it truly takes to adopt AI in manufacturing and achieve measurable impacts in operations. Key topics will include: - High-impact use cases that enhance throughput, quality, and Overall Equipment Effectiveness (OEE) - Incremental, scalable implementations that align with your existing roadmap - Governance models that effectively balance speed, cost, and risk Participants will learn how AI, digital twins, advanced analytics, and generative AI are currently being utilized to reduce lead times, enhance quality, and drive innovation across production, maintenance, and industrial operations. Date: May 27, 2026 Time: 9:00 AM – 1:00 PM (Lunch provided) Location: Automation Alley, 2675 Bellingham, Troy, MI 48083 Register here: https://lnkd.in/exDJMCi6 Featured speakers include: - Jason Hamp – Enterprise AI Technologist, NA Manufacturing, Lenovo - Jonathan Wu – Global Manufacturing Vertical CTO, Lenovo Manufacturing leaders looking to transition from AI experiments to real, repeatable operational impact are encouraged to reserve their seats using the registration link above. #AIinManufacturing #SmartManufacturing #IndustrialIntelligence #FactoryFloor #DigitalTransformation #Industry40 #Lenovo #NVIDIA #AutomationAlley #Integr8What does it take to move from AI experimentation to real operational impact? | Automation AlleyWhat does it take to move from AI experimentation to real operational impact? | Automation Alley
-
Peter Orbán shared thisJust heard something that reframes agentic commerce conversation. Luke Gebb, Global Head of Innovation at American Express, made a point that stuck: once AI agents start making purchases on your behalf, they become a new error source in the payment chain. And the existing chargeback model wasn't built for that. His framing: KYC becomes KYA. Know Your Agent. The implication is significant. AMEX is already thinking about agent purchase protection, absorbing losses caused by agent mistakes rather than leaving the customer exposed. That's not a product feature. That's a new liability category. And it means the payment infrastructure we've built over 50 years has to evolve before agentic commerce can scale with any confidence. Not sure we're ready. But the right people are asking the right questions. And the smartest of them I know in this space is Erin McCune #ReutersMomentum
-
Peter Orbán shared thisAaron Levie CEO of Box, said something yesterday at #ReutersMomentum that stopped me mid-note. "You use AI a little bit and you're like, oh my God, it's going to automate the entire business. And then you use AI a lot and you're like, oh my God. We need humans for almost every one of these steps." Two days at Reuters Momentum AI in New York. Two days of conversations with CIOs, CTOs, and CDOs from some of the largest enterprises in the world. Aaron's line captures exactly what I kept hearing in the hallways. The gap isn't between believers and skeptics. It's between people who've done one pilot and people who've done twenty. The ones who've done twenty aren't less excited: they're more precise. They know which steps agents can own, which steps need a human in the loop, and which steps nobody has figured out yet. That precision is what separates AI programs that compound from ones that stall.
-
Peter Orbán liked thisPeter Orbán liked thisEmber 1.0 came out on Aug 31, 2013. The first version of my book followed about a year and a half later, on February 18, 2015. I didn't mean to time it that way, but today is the day I'm making "Rock and Roll with Ember" free, exactly 13 years after the framework's first stable version. This has been an amazing journey, and I want to say thank you to all who were part of it ❤️ I wrote up a bit of history in a blog post that you'll find in the first comment below.
-
Peter Orbán liked thisPeter Orbán liked thisJust a little gratitude & picture sharing post. Thankful to still be doing what I love, with an immensely impressive team of humans. Not to mention... Switzerland 🇨🇭!
-
Peter Orbán reacted on thisPeter Orbán reacted on thisGoodbye Augment IT, Sometimes a new chapter requires a jump into cold water. Similar to one of my first days at Augment IT: jumping into the Aare in Bern with the team, not closing my wet bag properly, and ending up with all my clothes completely soaked, including the shoes I had to wear to a customer workshop the next day. A rather special, but unforgettable “onboarding”.😊 Now, six years later, I’ve decided to leave Inspect AR by Augment IT and start a new chapter. Behind me are intense years in Sales, the carve-out and building of an independent company as co-founder, and many exciting customer projects in the space of digital transformation, XR and AI. Thank you to all our customers for the trust, to RealWear and NavVis for the strong partnership, and to Reto Grob, Zdravko Nikolovski and Andrej Vckovski for the many years together, working on strategic topics, shaping the company’s direction and vision. And of course, thank you to the whole team in Switzerland and Macedonia for the teamwork, the expertise, the team meetings and all the get-togethers, from boat rides on Lake Ohrid and Lake Zurich to dinners and drinks. These are the moments that really bring a team together. For me, it’s now time to take a break, spend some time with my family and recharge by the sea. And in October, it’s time to start a new chapter. Stay tuned.
-
Peter Orbán liked thisPeter Orbán liked thisWhen my family moved to Washington, DC three years ago, I was not entirely sure what the next chapter would hold. Looking back, I'm grateful for what it became. I founded EnvisionIT as my own technology consulting practice. Supported Guidant with their ERP strategy and broader ERP journey. Joined the board of the SIM - Capital Area Chapter (SIMCAC), a community that quickly became one of the most meaningful parts of my time here. None of it was a straight path, but I've always listened to my instincts, and once again, it came down to saying yes to good people and hard problems. Now I'm ready for what's next. I'm happy to share that I've recently joined Fluence as Senior Director of IT, Sales & Corporate Applications. Energy storage is moving beyond simple backup to become a key enabler of modern infrastructure, and Fluence sits right at the intersection of renewable energy, the explosion of data centers, and AI. The company is scaling fast, with Siemens and AES in its lineage, which means the enterprise technology has to grow into what the business is becoming rather than catch up to it. I'm excited about the opportunity to work on it with a team I already respect. To everyone who made these last three years what they were, thank you! This next chapter is built on it.
-
Peter Orbán liked thisPeter Orbán liked thisWe're delighted to welcome Lenovo to the #SmartManuAuto program, where the team will share insights across multiple sessions exploring the future of AI-powered manufacturing. On 10 September, Jason and Jonathan will contribute to a presentation, roundtable discussion, technology demonstration, and panel session, bringing real-world perspectives on how manufacturers are moving beyond AI experimentation to achieve measurable business outcomes. A key highlight will be Lenovo's presentation, "When Innovation Becomes Operational: Lenovo's Lighthouse Factories Lead the Way", showcasing how AI-driven manufacturing is delivering reduced lead times, lower logistics costs, fewer quality losses, and higher operational efficiency. Their roundtable, “From AI Pilots to Manufacturing Impact: Scaling Computer Vision, Physical AI, and Agentic AI", will discuss the realities of taking AI initiatives from proof of concept to enterprise-wide deployment. The conversation will explore practical applications of computer vision, intelligent automation, robotics, and agentic AI, alongside the governance, infrastructure, and organizational readiness required to scale successfully. From smart factory innovation and operational excellence to the next generation of intelligent manufacturing, Lenovo's contributions will provide valuable lessons for organizations looking to unlock greater value from their AI investments. We look forward to welcoming the Lenovo team in Detroit and hearing how they're helping shape the future of manufacturing through AI, automation, and digital transformation. Register here: https://lnkd.in/dqt-FptQ #ManufacturingAI #Industry40 #Lenovo #ComputerVision #AgenticAI #DigitalTransformation #AMGWorld
-
Peter Orbán liked thisPeter Orbán liked thisAs I mentioned in my previous post, Jeff and I at AfterNow have been contracting with Microsoft on fundamental research for the past three years—research that, until now, has been confidential and closely guarded. Let me share this project with you. It began with using light to enable servers to communicate across racks, replacing cables with small projectors and data matrices. Then Teresa Nick applied neuroscience to develop a concept for mathematical operations using additive light on projections for GPUs during AI training. During AI training, GPUs need to sync and share data with each other—a step called an AllReduce operation. This is both a bandwidth and performance bottleneck, and it consumes significant energy. Teresa's idea: project GPU data as additive light onto a single surface and capture it back with the GPUs. This would execute the AllReduce operation instantaneously. It's called Parallel Photonic Integration (PPI). Jeff and I built a simple prototype using four GPUs, keeping key constraints in mind for future performance—everything needed to run in GPU memory, from projection to camera capture. It was an amazing challenge. The prototype worked and proved the overall logic and architecture were sound. We were on our way to building a second prototype with six servers and 48 A100 GPUs, complete with custom optics systems, to validate energy consumption and potentially demonstrate speed superiority over existing solutions. Unfortunately last year, Microsoft shifted priorities and the funding was diverted. For the past year, we've continued working on the paper and improving the prototype's performance and capabilities through a partnership between UCSD and Microsoft. I am very proud to announce that this month our paper on PPI was accepted in Nature Communications. https://lnkd.in/g2T_Z9dz This result is thanks to Teresa Nick's relentless work against all objections, as well as Jeff Breugelmans, PhD, Shane Drexler, Nikhil Jain, Hakki Refai, and with the support Husam Alissa. As the PPI prototype has evolved and improved, watching an AI learn has become quite visually fascinating. We look forward to bringing this technology to the next level—inside data centers.
-
Peter Orbán liked thisPeter Orbán liked thisI have enough computers running AI agents (Hermes) that the fleet needed its own map. Here it is. Mercury is in the middle. It is my principal engineer and the always-on controller for the homelab. Monitoring. DNS. Orchestration. Shared agent state. The boring work that makes the fun work possible. Jupiter and Saturn are the GPU systems. Venus is Grace Blackwell (it's a spark). Earth, Pluto and Voyager handle interactive work and desktop automation. Ganymede serves the working data. Callisto protects the backups. Other Hermes agents live across the fleet too, with their own machines and jobs. Mercury keeps the system coherent and tells me when I am about to do something questionable. Mercury also drew this. Version one was cursed. I told it so. It rebuilt the diagram and stripped the private details before I put it here. Honestly, that is a pretty good summary of how we work.
-
Peter Orbán liked thisPeter Orbán liked this📢 Join us at MATCH XR 2026 in Helsinki! We're excited to share that Mark Sage, Executive Director of the AREA, will be speaking at MATCH XR 2026, one of Europe's leading events for the XR ecosystem. 📅 November 12, 2026 📍 Helsinki, Finland MATCH XR brings together creators, enterprise leaders, investors, startups, and technology innovators for a day of visionary discussions, live XR demonstrations, and curated networking opportunities. It's a unique opportunity to connect with the people shaping the future of immersive technologies and enterprise adoption. Whether you're developing XR solutions, investing in the next generation of technologies, or exploring practical enterprise applications, MATCH XR offers valuable insights and connections. We look forward to seeing the XR community in Helsinki! Register today: https://matchxr.fi/ #MATCHXR #XR #ExtendedReality #EnterpriseXR #AugmentedReality #VirtualReality #Innovation #TechEvents #AREA
-
Peter Orbán liked thisPeter Orbán liked thisProud to announce that Casper Studios is being acquired by Ode with Anthropic! We started Casper as a curiosity. We wanted to learn about AI and see how we can use the technology to help companies, people, and ourselves. It’s since turned into something more. Our ambition has grown. And that is why we’re being acquired by Ode! Ode is the joint venture started earlier this year with Anthropic, Blackstone, Hellman & Friedman, Sequoia Capital, and others, led by Chris Taylor and Eddie Siegel. We’ve looked up to Chris and Eddie since we started Casper. Chris and I would trade notes about AI services. I always admired their energy; they are some of the best technical folks I’ve met. We joked that we should maybe join forces at some point. Can’t believe that’s now happening. What’s the high-level thesis on the partnership? Ode is acquiring Casper to expand its ability to support clients’ AI transformation from end to end. Ode goes deep on its clients’ top AI priorities, building production-grade applications for an organization’s hardest AI challenges. Casper goes broad, helping deploy AI across teams and repeatable workflows, mainly on harnesses like Claude Chat, Code, and Cowork. Following the acquisition, Ode will be able to address more of the customer opportunity without moving away from the complex, custom work at the core of its business. Thank you so much to the Casper team, all our clients, and all the supporters in our audience who have connected us with people and helped our company reach this milestone. It feels like the constraints have now been lifted. The future is quite literally unbounded in terms of opportunity. So excited about what comes next.
Experience & Education
Licenses & Certifications
Publications
-
A New Division of Labor: IoT, Wearables and the Human Workforce
AR for Enterprise Association Blog
There has been some consternation about the impact of automatization, frequently bordering a bit on sensationalism: AI (embodied or otherwise)"... Will Take Over The World!" But there is an increasingly clear picture that - at least for the foreseeable future - AI is a great fit for certain things but not so much for others. Notably: empathy, autonomous decision making, or - almost anything related to - emotions ... human folks are still better at it. So the emerging picture is about a new…
There has been some consternation about the impact of automatization, frequently bordering a bit on sensationalism: AI (embodied or otherwise)"... Will Take Over The World!" But there is an increasingly clear picture that - at least for the foreseeable future - AI is a great fit for certain things but not so much for others. Notably: empathy, autonomous decision making, or - almost anything related to - emotions ... human folks are still better at it. So the emerging picture is about a new division of labor between the human and artificial workforce mitigated and catalyzed by a new device class: #Enterprise #Wearables. Keith and I wrote up the vision we've developed together based on our first-hand experiences in the current IoT and Mixed Reality practice. #iot #digitaltransformation #mr #industry4_0 #vr #wearabletech #ar #ai
Other authorsSee publication -
Discovering the Value of Earned Audience - How Twitter Expressions Activate Consumers
FOX Broadcasting; Twitter; Advertising Research Foundation
Fox Broadcasting Company (FOX), Twitter and The Advertising Research Foundation (ARF) today announced new research exploring the value of the earned audience that media companies and brands gain from TV and brand-related tweets. Using Twitter’s capability to survey users on Twitter.com and the Twitter mobile app via homepage recruitment, the study, “Discovering the Value of Earned Audience – How Twitter Expressions Activate Consumers,” uncovered that:
Other authorsSee publication
Courses
-
Six Sigma Greenbelt
-
Projects
-
Data Readiness for the Agentic Enterprise: Preparing Your Organization for Agent-Native Operations
-
The session goes deeper on a dimension that most agentic AI conversations skip entirely: data readiness. AI agents do not consume data the way human analysts or dashboards do. They make tool calls, perform structured extraction, retrieve context across systems, and act on what they find, often in multi-step workflows with no human in the loop. Most enterprise data architectures wer designed for human-driven queries, not agent-driven operations.
This session explores what agent-ready…The session goes deeper on a dimension that most agentic AI conversations skip entirely: data readiness. AI agents do not consume data the way human analysts or dashboards do. They make tool calls, perform structured extraction, retrieve context across systems, and act on what they find, often in multi-step workflows with no human in the loop. Most enterprise data architectures wer designed for human-driven queries, not agent-driven operations.
This session explores what agent-ready data; looks like: how data structures, metadata, access patterns, and quality standards must evolve as agents become the primary consumers of enterprise information. The parallel to the SEO-to-GEO transition is useful here: web content had to be restructured for AI search engines, and enterprise data faces the same shift. We look at the practical steps regulated enterprises should take now to prepare internal datasets, knowledge bases, and data
governance for an agent-native future. -
NYC Leaders in AI Summit - Agentic AI workshop
-
AI Deployment Workshop – Part 1
Sovereign AI in the Age of Agents:
Architecture, Control, and the Economics of Autonomy
Sovereign AI has become one of the most discussed concepts in enterprise technology. But what does sovereignty actually mean when your AI systems are autonomous agents? Agents consume tokens at 15-50x the rate of chatbots, make decisions regulators will scrutinize, and interact with your enterprise data in ways most architectures were never designed for. This…AI Deployment Workshop – Part 1
Sovereign AI in the Age of Agents:
Architecture, Control, and the Economics of Autonomy
Sovereign AI has become one of the most discussed concepts in enterprise technology. But what does sovereignty actually mean when your AI systems are autonomous agents? Agents consume tokens at 15-50x the rate of chatbots, make decisions regulators will scrutinize, and interact with your enterprise data in ways most architectures were never designed for. This workshop presents a five-dimension framework for sovereign AI in the agentic context: compute sovereignty, token sovereignty, data sovereignty, decision sovereignty, and operational sovereignty.
Each dimension maps to specific architecture decisions, from right-sizing across the full infrastructure stack to preparing enterprise data for agent-native consumption. Grounded in token economics analysis showing up to 18x cost advantages through infrastructure ownership and real-world deployment architecture from regulated industries, this session gives enterprise leaders a practical framework for maintaining control as agentic AI moves into production environments.
Languages
-
English
Native or bilingual proficiency
-
Hungarian
Native or bilingual proficiency
-
Russian
Limited working proficiency
-
German
Elementary proficiency
Recommendations received
16 people have recommended Peter
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Peter Orbán
72 others named Peter Orbán are on LinkedIn
See others named Peter Orbán