Lights Out: Technology and Automation Podcast by MakingChips

Most automation conversations stop at the cell. This one keeps going, through a fab shop, a machine shop, a Department of Defense contract, and eventually a coal mine. Michael McHale runs PSA Systems, an automation integrator in Pennsylvania that also fabricates, machines, and delivers turnkey projects in house. He bought the business in 2016 with four employees and about $900,000 in revenue. Today it's 100 people, 65 engineers, and over $30 million, split across five verticals that all feed each other. That structure is the whole point of this episode. When PSA chases something that looks like a distraction, like reclaiming and selling coal, it works because they already own every capability that business would otherwise have to outsource. We also get into the part almost nobody wants to talk about, which is time. Mike runs on a five year plan and makes capital commitments for 2029 and 2030. He bought an $800,000 five axis mill that was closer to a million by the time it was wired and sitting on concrete, and he committed to it six or seven months before he had work to put in it. He's six figures into an IBM Watson rollout that he openly says won't show real value for another three to five years. His point is to stop treating technology like a series of purchases and start treating it like a position on where the business is going. There's a practical thread for smaller shops too. Mike doesn't think a 15 or 20 person shop needs an enterprise AI rollout, and he says so. But if you're holding controlled information or working in regulated industries, he walks through why the guardrails matter, what goes wrong when drawings end up in a public model, and why the CMMC picture is so confusing right now. Then it gets personal. Broke, rich, double broke, way richer than the first time. We asked how you stay level headed on that ride, and his answer came back to people instead of EBITDA. Can your employees buy a house? Can they send their kids to college? That's the scorecard. What's Covered in this Episode (0:00) Broke, rich, double broke: the roller coaster of running an industrial business (0:41) Meet Michael McHale of PSA Systems, and what's driving automation demand in 2026 (0:41) Where the robots take our jobs argument breaks down, and why skilled headcount grows (0:41) Running an automation integrator and a machine shop under one roof, and the retirements pushing work to both (7:48) The capital arms race: a million dollars into a five axis mill before it makes a dollar (8:56) Hennig (10:08) Retrofitting the machines you already have instead of waiting on new iron (10:27) Tolerance and quality demands are separating who can serve which customers (11:35) Chasing EBITDA now versus building for 2030, and running on a five year plan (13:54) Automation assessments: pricing one project versus walking the whole plant (14:31) Why the US C-suite wants a two year ROI, and how Europe and Japan think about it (17:26) Navu (18:42) What actually triggered the IBM Watson rollout, starting with project management (20:52) From a static agent to an active one, and moving documentation into Jira (24:52) A six figure AI bet that pays off in 36 to 60 months, not 12 (26:26) Hire MFG Leaders (27:54) DIY AI with Claude or Grok versus an enterprise rollout with guardrails (27:54) Why a DoD contractor holding customer IP can't put drawings in a public model (31:38) CMMC confusion, the pause, and what smaller regulated shops have to weigh (32:51) From four employees and $900K in 2016 to 100 employees and over $30 million (35:16) Why PSA built verticals when most integrators still haven't (38:33) Buying a coal mine, and why it fits a vertically integrated industrial (40:06) CLA (43:28) Hitting the accelerator at the bottom of the dip instead of coasting (46:20) The real scorecard: can your people buy a home and send their kids to college Resources Mentioned PSA Systems Hennig WorkFlow Navu Hire MFG Leaders CLA IBM watson Fastems Slytek Connect with Michael McHale Michael McHale on LinkedIn PSA Systems on LinkedIn PSA Systems

When Uriel Eisen needed more capacity in his 325-square-foot shop, his first instinct was to buy a robot. But the machine wasn't the real bottleneck. Production kept getting derailed by missing material, forgotten orders, and inexpensive supplies that nobody noticed until they were gone. Uriel's path from designing spacesuits to manufacturing CNC-machined hardware eventually led him deep into Kanban, lean manufacturing, and continuous improvement. What began with index cards and a Sharpie ultimately became Arda—a system built to connect what happens on the shop floor with the people responsible for keeping it supplied. In this episode, we explore why some of the best automation projects look surprisingly low-tech, how small interruptions quietly consume far more capacity than most shops realize, and why a process that depends on discipline may actually be designed to fail. Before you invest in another machine or shave a few seconds from a cycle, take a closer look at what's already stopping the work. The fastest return in your shop may be hiding inside an empty bin. What's Covered in this Episode (0:00) The fastest ROI automation project on your floor has nothing to do with robots (1:03) Meet Uriel Eisen of Arda and Austere Manufacturing (4:18) Designing spacesuits for NASA, and why patterning fabric is like flat wrapping sheet metal (5:28) Why we love SMW Autoblok for workholding (6:12) How Uriel started designing for NASA (9:09) Meeting at the Machining Summit and the talk on inventory management (9:55) From sketching buckles to buying a CNC router, a Tormach, and a Speedio (11:47) Running out of aluminum and the three cent plastic bag that stopped a shipment (12:50) Toyota Production System, Paul Akers, and running all of it from a 325 square foot barn (13:15) Kanban doubled throughput, then doubled it again, with no extra space or people (13:57) What a bad week looked like when "let me know if we're running low" was the system (16:09) The flip side: overordering ties up cash and hides the real problem (17:27) The customer who scrapped five million dollars of obsolete cardboard boxes (18:05) Take your shop to the next level with DN Solutions (19:16) Overproduction and implementing SMED to cut setup time (20:59) Kanban cards piling up on the desk and how that frustration turned into Arda (24:08) Five minutes of training, five dollars of supplies, and 20 percent of floor time back (26:02) Reliability beats cycle time, and systemization beats your best firefighter (29:15) Shadow board inertia, color coding, and why the right way has to also be the easiest way (35:50) Build your pipeline and engineer your growth with Factur (36:55) Paper cards in 2026, the sexy robot problem, and why the world already converged on pull (42:11) How to start on Monday (and what to do before you ever call Arda) (44:27) Give your people an hour a day to improve their own work and watch throughput double (48:50) The takeaway: automation doesn't always look like a robot arm Resources Mentioned Arda Cards Arda Free Kanban Card Generator Austere Manufacturing SMW Autoblok DN Solutions Factur Hennig IMTS Connect with Uriel Eisen Connect with Uriel on LinkedIn Incremental: The Continuous Improvement podcast Arda on Instagram Austere Manufacturing on Instagram Arda on LinkedIn

Eric opens this episode describing a dream he had. A shop with no standalone machines. Three hundred and nine pallets moving 1,600 times a day. A central tool system holding 5,500 tools, five Fanuc robots shuttling up to 800 of them daily, and not a single crash. The punchline is that it wasn't a dream. He'd just walked through it. That shop is Norbert Kempf CNC-Technik in St. Ingbert, Germany, a 100-person contract machine shop that runs simple to highly complex parts in batches of one to 500. We caught up with owner and managing director Stefan Kempf on his own floor, along with Risto Niemi of Fastems, who has spent 28 years helping shops rethink how they grow, and Yanick Müller of swissMODULAR, who thinks about work holding the way most people think about religion. The number that stops you is this one: Setup time inside the machine is zero. Every tool arrives staged, every pallet is built offline, and the price a customer pays is the same whether they order ten pieces or a thousand. Utilization sits above 90 percent every single day, across 23 four-axis horizontals split between two buildings and joined by a bridge with a robot running tools across it. Any machine can run any part, so there is no bottleneck to schedule around. What makes this more than a hardware story is how Stefan got there. He spent three years preparing the business before the first flexible manufacturing system arrived, standardizing machine brands, tool interfaces, and pallets into what he calls a monoculture of machines. Then he built a bonus system that pays the whole company on utilization and measures it every shift, so a bad night is just a bad night and tomorrow is a fresh chance. If you've added machines and watched your utilization fall, this is the conversation to sit with. Stefan, Risto, and Yanick each leave you with one tactic you can act on, and none of them require a new building. What's Covered in this Episode (0:00) Eric's dream: a shop with no standalone machines, 309 pallets, and not one crash (0:58) Inside Norbert Kempf CNC-Technik in St. Ingbert, Germany, and its CNC 4.0 concept (6:48) Risto Niemi of Fastems on 28 years of helping shops rethink how they grow (8:53) Stefan Kempf on taking over his father's 1970 turning shop and building what stands today (10:21) Yanick Müller of swissMODULAR on selling a philosophy, not just a vise (12:02) 1990 to 1999: one machining center becomes ten, and utilization goes the wrong way (13:18) Monoculture of machines: standardizing brands, tool interfaces, and pallets (13:54) Why Kempf builds its own machine pallets and now sells them worldwide (15:25) MakingChips, DN Solutions, and Kennametal are giving away a fully tooled SVM 4100 (16:09) The frustration that started it: walking in to the same errors every single day (18:00) Risto on why buying another machine is the easy answer and usually the wrong one (19:24) Yanick on using the full envelope you already paid for (20:58 A decade-long vision and three years of preparation before the first FMS (23:19) Inside the CTS: five Fanuc robots, 23 horizontals, tools staged before the cycle calls (24:41) The IMTS Industrial AI Conference, September 16 in the South Building (25:29) Zero setup time inside the machine, and what that does to a quote (27:37) How good business relationships help makes everyone stronger (29:42) The bridge between two buildings and the robot that runs tools across it (32:27) Tiny parts, big parts, 100 on a tombstone, and why every machine runs every job (35:00) One operator for two machines, one troubleshooter for all 23 (36:07) Navu adds AI chat to your website, trained on your own content (37:19) Multi-skilled people, the Swiss apprentice system, and unmanned nights (41:03) A bonus system tied to utilization and measured every shift (46:07) One tactic each: tool management, spindle up time, and never stop learning Resources Mentioned Norbert Kempf CNC-Technik GmbH Fastems swissMODULAR MakingChips Machine Giveaway IMTS Navu Hennig Connect with Stefan Kempf, Risto Niemi, and Yanick Müller Norbert Kempf CNC-Technik GmbH Fastems swissMODULAR

What if the most powerful sensor on your shop floor isn't a camera or a probe or a vibration puck—but a power cord? That's the premise Lauren Dunford, co-founder and CEO of Guidewheel, brought to the show, and she explained it using a hundred toasters. Guidewheel clips a current transformer around the cord and reads the electrical heartbeat of a machine. No PLC integration, no open heart surgery, no waiting on a controller vendor. From those tiny fluctuations the platform gets run, idle, and down, then cycle counts, cycle times, early maintenance warnings, and in a lot of cases which SKU is running and where the anomalies are. It works on your CNC equipment, and it works on the compressors, chillers, conveyors, and palletizers that quietly stop your plant. We also got into the question this show keeps circling. Is monitoring surveillance? Lauren's answer is that the data has to be used with the team, never on it. She would rather see posters, t-shirts, and a shift competing to win the hour than one manager holding a dashboard nobody else can see. If you have ever suspected your utilization number is fiction, this one is worth your commute. What's Covered in this Episode (1:30) The most powerful sensor on your floor might be the power cord (6:03) Lauren's origin story: a hairnet, a food plant, and being hooked for life (7:36) The parking lot calls to customers that sent her back to Stanford (8:34) Rebuilding her product around how the customers used it (9:51) Shipping a rough prototype and letting three VPs of Ops tear it apart (13:23) What you should never vibe code: PLCs, safety, and data integrity (14:24) MakingChips machine giveaway with DN Solutions and Kennametal (15:09) Slytek in the cut versus Guidewheel across the whole plant (18:58) Heartbeat of everything: From run, idle, and down to part counts and cycle times (21:13) Spotting which SKU is running and flagging cycle anomalies for quality (21:49) Just because you can doesn't mean you should (24:06) Why we love SMW Autoblok for high-quality workholding (24:50) The anomaly alert that caught smoking machines before a fire (26:48) 42% to 95% uptime from staggered breaks, on-time starts, and microstops (28:39) The opportunity hidden in "unsexy" things in the industry  (29:45) The rage quit, five waiting on tool events, and a 37 minute payback (30:33) Real utilization by industry and the money left on table (32:43) Invest in ProShop ERP to invest in your business (34:19) The CapEx request for a machine that was actually running at 7% (38:16) Building a culture of continuous improvement and growth mindset (44:04) Is monitoring surveillance? Using data with the team, not on the team (48:01) Lauren's pitch to a 22 year old considering manufacturing (50:20) Three tactics: get good data, turn it into action, and listen to your team Resources Mentioned Guidewheel Slytek Machine Intelligence MakingChips Machine Giveaway IMTS SMW Autoblok ProShop ERP The Goal Connect with Lauren Dunford Guidewheel Connect on LinkedIn

Every shop runs on a little tribal knowledge, the person who just knows where that tray of parts is headed and where it tends to get stuck. It works right up until they take a day off and the whole floor is hunting for a job that vanished somewhere between the mill and the heat treat rack. That gap is exactly what this episode is about: what happens when you stop guessing and start actually seeing where everything is. Dave and Eric brought in Mitch Tucker of HID to break down RTLS, real-time location systems. In plain terms, RTLS tracks where your things are, your parts, your tooling, your material, even your people. The part that matters for us is where that data lives. It sits in the white space between your controlled processes, the handoffs and the staging and the waiting, and that's where most shops are quietly bleeding time. Mitch's big point is that this isn't just a giant-warehouse toy. The same search-time problem that costs a 200 person facility around $400,000 a year shows up in a five person shop as a machine sitting idle while somebody looks for a tool that's thirty feet away. He makes the case for starting small and scaling deliberately, from a simple RFID setup to Bluetooth to ultra-wideband, matching the system to the size of your business instead of buying the most expensive thing in the room. There's a real consulting streak to how Mitch approaches this, and it lines up with everything we preach. You don't drop RTLS into a chaotic shop. You know your processes first, you go to GEMBA and watch the work, you do your 5S, and then you add the visibility layer on top of a floor you already understand. From there the location data starts paying off in places you don't expect: the connected worker, safety and evacuation roll calls, job costing, and the honest conversation about where useful data ends and surveillance begins. The throughline is one we come back to constantly. You can't run lights out, and you definitely can't jump to what Mitch calls Industry 5.0 and the AI wave behind it, without the data foundation underneath. Location data is a huge part of that foundation. If you've ever lost an afternoon to a missing tool, start here. What's Covered in this Episode (0:00) A real-world RTLS use case: intruder detection at an aerospace and defense facility (3:35) Learn more about Mitch Tucker and his role at HID as a consultant (5:00) What is RTLS? How is it useful in manufacturing?  (6:20) The Kent problem: tribal knowledge and the parts that get lost between processes (8:03) Mitch's path from engineering drawings to continuous improvement evangelist (13:26) Is RTLS only for giant warehouses? Who does it apply to? How does it work? (18:56) Head to Kennametal's Next Level Shop at IMTS 2026, booth 431800  (20:00) Digital twins, RTLS data, and using real-time visibility as a sales tool (22:20) Why RTLS is becoming a commodity and the coming rush toward Industry 5.0 (24:44) A wild use case revisited: cameras, LIDAR, and location data for intruder detection (26:22) Why you should head to the IMTS Job Shops Workshop (27:13) How to actually get started and why the consultant comes before the sale (30:27) Scaling from RFID to Bluetooth to ultra-wideband, and installing once (33:24) Rolling RTLS from one facility to the next on the same software (34:20) The $400,000 question: what search time quietly costs a shop (37:36) Thin margins, throughput, and how one lost tool can sink a small shop (39:53) Connected worker, safety, and job costing without turning into Big Brother (44:08) Why we created Hire MFG Leaders and why you should use it (44:37) Mitch's low-hanging fruit: know your processes, go to GEMBA, and do your 5S Resources Mentioned HID Kennametal IMTS Job Shops Workshop Hire MFG Leaders Connect with Mitch Tucker HID Connect with Mitch Tucker on LinkedIn