There’s a certain kind of conference moment you don’t see coming until you’re already in it. Not a flashy product demo or a jaw-dropping stat. A theme that keeps surfacing, session after session, until you can’t ignore it anymore.
That’s what happened at AMS.
If you walked in expecting wall-to-wall AI hype, you’d have found some of that. But the loudest signal - from keynotes to breakout sessions to hallway conversations - wasn’t “look at this new thing.” It was: get the foundation right first.
The companies saying it weren’t scrappy startups or cautious laggards. They were 3M, Toyota, Medtronic, HelloFresh, and others - household names with serious operational scale. And they were saying the same thing, in different ways, on stage after stage.
“We amplified a bad system”
Rebecca Teeters, 3M's SVP of manufacturing operations, opened her session with a confession that probably landed differently for different people in that room.
“We drifted from our first principles,” she said. Constant organizational change, short-term pressure, talent turnover - the fundamentals slipped. And when they layered technology on top anyway, they didn’t go faster. They went faster in the wrong direction. “We amplified a bad system.”
The room went quiet.
Her point wasn’t to slow down on technology. It was more specific than that: “The implementation of the technology becomes the goal versus the amplification of the first principle.” Technology isn’t the problem. The sequence is. 3M is now doing the foundational work first - before layering on anything new.
She also said something worth sitting with: “Operational excellence is not complicated. It’s just not. It’s very hard. Very difficult to get right day in, day out. But it is not complicated.”
Hard vs. complicated. A distinction that gets lost quickly.
Lean first. technology next. people always.
Kelly DiPuccio, Medtronic's digital manufacturing leader, built her entire session around one line: “Lean first, technology next - and people always.”
She named a pattern that gets expensive fast: “random acts of automation.” Deploying technology onto a process that isn’t stable. Chasing an implementation milestone instead of an operational outcome. “Sometimes we would deploy automation on a process that wasn’t stable.” The result wasn’t improvement - it was accelerated chaos.
Their approach now: establish the lean foundation, prove process stability, then bring in technology to scale it. Medtronic runs a platform processing 260,000 images per day across 13 AI models simultaneously - but that only works because the process underneath it was built right first.
“We aren’t just deploying technology for technology’s sake.” In a room full of off the shelf technology vendors, that line shouldn’t be remarkable. But it kind of is.
Lean didn’t change. The speed did.
The most energetic session came from HelloFresh’s COO for North America. Six months ago, he said, he thought AI was “just a coding tool that individuals use to speed up T-shirt sizing.” His view has since shifted - not because AI replaced his operating methodology, but because it turbocharged it.
“Lean didn’t change. But the speed at which we deploy and learn is at a level I would have never imagined.”
HelloFresh embedded AI agents into their lean manufacturing operations and the results are real: 94% reduction in critical pull time for middle-mile delivery, 25% labor savings through demand-driven planning, significant cuts in overproduction. They’re now piloting an “infinite menu” concept that would have been operationally impossible to manage without AI-assisted modularity across 622 different tools.
The lean principles didn’t get disrupted. AI gave them a velocity they couldn’t achieve manually. “Predictive pull” - anticipating downstream demand shifts in real time rather than reacting to them - isn’t a new concept. It’s a lean concept that AI can now execute at a speed no human system can match.
Execution latency: The real cost of inaction
Christopher Rencado, who leads digitalization at W.R. Grace, a specialty chemicals company with 20 manufacturing facilities, kept his session focused on one specific, expensive problem. Their talk wasn’t about transformation strategy or platform architecture. It was about a specific, expensive problem: the gap between seeing an issue and doing something about it.
He called it “execution latency.” And he framed it as the defining cost challenge in industrial operations right now.
The picture he painted was familiar to anyone who’s spent time on a plant floor. A worker sees a problem. Starts walking to a desk to log it. Gets interrupted by three other operators with three other issues. By the time they sit down, they’ve forgotten what they were going to enter. The issue sits. Parts aren’t ordered. Work orders don’t get triggered. Equipment keeps running - or stops.
“Execution latency is the new cost of inaction.”
Their fix - a unified mobile platform connecting operations, maintenance, and stores teams - produced results that are hard to argue with. At their Chattanooga facility: 8x improvement in executed high-priority work orders, a 48-point jump in scheduled compliance, and 35% of maintenance work now coming from operator-initiated notifications (previously near zero). They also freed up 48% of previously unutilized capacity.
The technology isn’t complicated. The discipline to implement it consistently, meet workers where they are, and make the new way easier than the old way - that’s what takes work.
Toyota: Same foundation, digital speeds
Chris Nielsen, Chief Supply Chain Officer for North America at Toyota, oversees a $50 billion supply chain and 70 billion parts per year. COVID broke things in ways they didn’t see coming - two siloed supply chains, no upstream visibility, dealers making promises to customers they couldn’t keep.
“We were not good enough,” he said.
What they’ve built since is impressive: a unified data platform with real-time visibility across the entire supplier-to-dealer network. AI replaced 75 Excel spreadsheets in demand and supply planning. Digital twins let the team run scenario analysis in hours that used to take a full week.
The foundation? Still the Toyota Production System. The same lean principles that have guided the company for decades. TPS didn’t get replaced by AI. It got a new engine. “TPS operating at digital speeds with real-time data.”
Toyota’s transformation wasn’t from their production system. It was an expression of it, at a new scale and speed.
From existential disruption to operational amplifier
Here’s what nobody said explicitly at AMS, but what all of it implied together: the last few years of AI exploration may have been necessary. But they were also pretty scattered.
Fear of being left behind pushed a lot of organizations into AI-first thinking before the process work was done. Pilots launched and stalled. CFOs got disappointed quarter after quarter. Technology landed on systems that weren’t designed to support it. You can’t scale a broken process - you just break it faster.
A few years ago, most people at a conference like this fell into one of two camps: blind enthusiasm for AI, or quiet fear of it. AMS felt different. Those camps have mostly dissolved. What’s taken their place is something more useful: AI as a tool. A genuinely powerful one. But a tool in service of a strategy and a set of principles that actually work.
The HelloFresh COO who six months ago thought AI was just a developer productivity shortcut is now using it to run predictive lean operations at scale. Medtronic is running 13 AI models simultaneously - in service of a quality system that is fundamentally human-centered. Toyota didn’t rebuild on a new methodology. They rebuilt the Toyota Production System at digital speeds.
That shift - from “AI as existential disruption” to “AI as operational amplifier” - is the real story of the conference. And it’s a healthier place for industrial leaders to operate from.
We see this pattern regularly in our work with industrial clients. The organizations that get the most out of technology investments aren't the ones who moved fastest -- they're the ones who did the unglamorous work first. Defined the process. Stabilized the foundation. Got their teams aligned. The technology came after, and it landed. That sequence is harder to sell internally than a pilot. But it's the one that actually produces results you can defend to a CFO.
What this means if you’re an industrial leader
The message from AMS isn’t complicated. It might even sound familiar.
Get your foundation right. Know what your system is designed to do. Make sure your people understand the work before you automate it. Treat execution latency - the gap between insight and action - as a first-order cost problem, not an IT problem. Then use technology, including AI, to go faster, see clearer, and scale what actually works.
The tools are genuinely better than they’ve ever been. The AI is real. The results are real. But the best companies in that room didn’t get there by chasing the tool. They got there by earning the right to accelerate - by doing the hard, unglamorous work of getting the basics right first.
That’s not a new idea. It never was. But it’s good to hear it said out loud again, by people actually doing it.
About the author
Adam Schanfield is a Principal Strategist at TXI. He spends most of his time helping industrial companies figure out what's actually worth building -- and what isn't. That means a lot of conversations about data foundations, organizational alignment, and why good pilots don't always become good products
