50 - Why Factory AI Fails Without a Data Foundation with Alexander Kruger
Unplugged: An IIoT PodcastShow Notes
Most industrial AI projects fail before they produce anything useful. The models are ready, the dashboards look great, and the executive presentation landed. But on the shop floor, nobody can get clean data out of the PLC without calling the vendor. The missing piece is a data foundation. Alexander Kruger, co-founder and CEO of United Manufacturing Hub (UMH), joins Phil Seboa and Ed Fuentes on this episode, powered by PLCnext Technology, to explain why factory AI stalls without the right infrastructure underneath it, why boring technology beats shiny platforms, and how open source tooling is putting data ownership back in the hands of the people who run factories. Key topics in this episode: Why most factory AI projects fail at the data layer before they reach the model layer How UMH went from systems integration with McKinsey to building open source data infrastructure The five-step process for connecting machines, modeling data, and serving it to applications Why Kafka and Postgres beat purpose-built IoT platforms for long-term reliability How containers and Kubernetes solve high availability problems OT has wrestled with for years The open source bet on forerunner power users who change organizations from within -------------------------------- This episode is proudly made possible by PLCnext Technology PLCnext Technology is the ecosystem for industrial automation consisting of open hardware, modular engineering software, a global community, and a digital software marketplace.
Learn more at: https://www.plcnext-community.net/news/synergy-edge-cloud/
-------------------------------- Connect with Alexander on LinkedIn: https://www.linkedin.com/in/alexander-krueger/ Connect with Phil on LinkedIn: https://www.linkedin.com/in/philseboa/ Connect with Ed on LinkedIn: https://www.linkedin.com/in/edfuentes/ Learn more about United Manufacturing Hub: https://umh.app
More from Unplugged: An IIoT Podcast
This episode is brought to you by Canary Labs, and Cirrus Link Solutions.
This episode is brought to you by Cirrus Link Solutions and Canary Labs. Cirrus Link: https://unpluggediiot.com/go/cirrus-link Canary Labs: https://unpluggediiot.com/go/canary-labs
Phil Seboa and Ed Fuentes sit down with David Ariens (Analytics For Industry / AVEVA) and Willem van Lammeren (Syensqo) to explore why IT/OT convergence fails — and what organisations actually get right when convergence works.
What changes when the assets you operate stop being centralised and predictable, and start being distributed, intermittent and weather-dependent? The old operations playbook breaks — and most digital programmes are still designed for the old one. Dr Hema Wadhwa, Associate, Data & Analytics | Asset Management and Performance at Aurecon, joins Phil Seboa and Ed Fuentes to talk about building digital asset intelligence for the energy transition. Hema's career is a live map of that transition: she started as a Floating Systems Engineer in global oil and gas, moved through subsea engineering and power generation operations, and now leads digital strategy and analytics for mixed energy asset fleets. Key topics in this episode: - The unplanned career arc — floating production systems and subsea engineering into renewable operations - "Are you here to take my job?" — what digital transformation actually feels like from inside the control room - Why the solution providers have to evolve, not the operators - One gas turbine versus a hundred wind turbines: why renewable fleets break the old operations model - No standard across vendor SCADA screens and UIs, and what that inconsistency costs - Why there is no AI magic without the data groundwork first — and why selling it that way isn't ethical - Cognitive load: lean frontline teams, PDF report fatigue, and tools that land as a burden - OT/IT convergence and the cybersecurity questions that come with opening it up - OEM warranties and whether asset owners really have access to their own asset data - Recording data when it changes, not on a fixed scan rate - One source of truth, different views per role — data presented so it guides a decision - Predicting failures months ahead: cranes, spares, and turning downtime into planned work - BESS as a trading platform, not just another renewable asset - Why the real cost driver is people, and why it only stacks up if it scales across assets