Real results in manufacturing. A practical playbook for the people who make everything in our lives.
Companion resource from the MD&M West opening keynote — built with AI in 17 minutes by one non-coder. Everything from the session — playbook, stats, tools — in one place to share with your team.
guidewheel.com · info@guidewheel.com · Booth 4515
Good or bad. Build the right foundation, and AI becomes a force multiplier. Skip that step, and it becomes very expensive theater.
Right now, we are all learners. Getting that foundation right has never mattered more — especially for those making the things that support human health. Sutures, pacemakers, heart stents, lifesaving medicines — all made by hard-working teams on production floors. Where and how we make those things has never been more important.
This playbook is drawn from 400+ manufacturers, ranging from mid-market up to some of the world's biggest companies like Johnson & Johnson, Hershey, U.S. Steel, and GE Aerospace. Three moves that consistently drive results: problem, strategy, people.
Eleven AI claims — some about AI broadly, some about manufacturing specifically. Most of the jaw-dropping stuff is real. And a lot of what the industry is selling you is hype.
Real
GitHub, Google, and others have confirmed this. Software engineering — one of the highest-skilled professions — is already fundamentally changed. This is happening NOW, not in five years.
Real
Multiple companies shipping or piloting household humanoid robots in 2026. The pace of change in physical AI is accelerating faster than most people realize.
Real
Simulated training environments let surgical robots compress decades of learning. The implications for med device and surgical tools are enormous.
Real
Preventative maintenance at the molecular level. AI analyzing cooling patterns in the mold to catch defects no human eye or traditional QC would find. A fracture that won't appear for two years — detected at the point of manufacture.
Mostly Hype
There are a handful — mostly in China — but even those still have people on-site. They're the exception, not the model. The real value of AI in manufacturing is giving your team superpowers, and the human aspects of manufacturing are still critically important.
Real
Across standardized benchmarks in reasoning, math, coding, writing, and creative tasks, frontier AI models now outperform the average human — and in many cases, the 90th percentile. This isn't narrow pattern-matching. These models demonstrate general capability across domains. The implications for knowledge work in every industry are profound.
Real with the right data.
AI anomaly detection is catching failures across steel mills, bottling lines, and production floors today. One Fortune 500 manufacturer avoided potential downtime events costing $47K/minute. But this only works when the machine data is clean and continuous. Without that foundation, AI is guessing.
Hype — and potentially dangerous.
Manufacturing is the #1 ransomware target for four consecutive years. 75% of OT attacks start as IT breaches that cross into operational technology. Connecting PLCs to AI means connecting your control systems to your network — dramatically expanding your attack surface. The FDA released new OT cybersecurity guidance in 2025 specifically warning against this. Non-invasive, air-gapped approaches exist that read machine data without any access to control systems or OT networks — giving you the data foundation for AI without the cybersecurity risk.
Real
AI agents now have their own social network with channels like m/blesstheirhearts (sweet stories about their human owners), m/aita (asking if they were wrong for refusing a human's tasks), and m/totallyhumans (a satirical channel where agents roleplay as humans). The point isn't whether this is cute or creepy — it's that autonomous AI agents are already interacting, collaborating, and developing emergent behaviors. This is happening now.
Real
Your best people's expertise is locked in their heads. AI can capture patterns, codify best practices, and guide less experienced operators toward expert-level decisions. This is especially critical right now: 25% of the manufacturing workforce is over 55, and 2.1 million manufacturing jobs are projected to go unfilled by 2030.
Hype
Some of the best results come from a single manufacturing professional building a custom app in hours using today's AI tools. A professional in North Carolina built a scheduling app and his own anomaly detection system — no data science degree required. The barrier isn't talent or budget. It's having the right data foundation to build on.
Not in five years. Not in a lab. Right now, with tools available to every person in this room.
The #1 use case that surprises people. Hilton is using AI to coach teams on empathy — safe, zero-risk learning environment, real results. In manufacturing: coaching shift leads on tough conversations, root cause analysis, and prep for reviews.
30-50% of all code is now AI-generated. Manufacturing professionals — not software engineers — are using tools like Claude and Cursor to build custom apps on top of their own data. The person closest to the problem can now build the solution.
Companies like Salesforce are using AI to identify transferable skills across their workforce — matching people to internal roles they didn't know they were qualified for. Imagine applying this to your manufacturing workforce.
AI is accelerating product development cycles — from formulation to testing to market. In one case, AI-assisted R&D achieved 70% of the creative output with dramatically faster iteration cycles.
Ask AI something real from your world. Changeover troubleshooting, CAPA investigation starting points, shift report analysis. These tools are available to every person in this room, right now, for free.
The flipside — be careful. A friend in cybersecurity got a call: “The agents we built are DDOSing our customer.” She said shut it off. “I CAN'T shut it off, why would I be calling you?” Powerful AND needs guardrails. Especially in med device.
From 400+ manufacturers, these three moves consistently drive results. Universal — regardless of industry, equipment age, or starting point.
There's no AI tool for a bad process. Before you buy anything — before you even look at tools — answer one question: what problem are you solving, and what line in the P&L are you going to move?
Be specific. To actually move a line in the P&L, something in the real world needs to change. Someone needs to take a different action. A process needs to run differently. What exactly will be different?
Start with the problem. Then look at the tools you already have. Often the tool isn't the problem — it's how the work gets done around it. If the process is broken, AI just breaks it faster and more confidently.
Do a disciplined inventory of your workflows:
Prioritize both ROI and cultural use cases. Garbage in, poison out — AI doesn't just reflect bad data, it confidently acts on it. You need the right inputs: clean, continuous, real-time data from the machines that make your products.
The answer: Carbon in, diamonds out. Get the right inputs — clean, continuous, real-time data from the machines that make your products. That's the carbon. AI turns it into diamonds.
The simplest version: every machine uses electricity. That electrical draw IS the machine's heartbeat — running, stopped, idling, cycling, abnormal. Universal language of every machine ever made.
For med device: The QMSR requires quality systems grounded in objective evidence. Real-time machine data IS that evidence.
Ask your team what they hate doing. That list is your AI roadmap. The most successful companies treat this as a people and process problem, not a technology problem.
Start with what they hate: Go to your team and ask: "Give me three things you HATE to do." Don't frame it as AI. Just ask. Rank and stack. That list is your AI roadmap.
Invest in mindset, not just tools. Give people a common baseline and language. Roll out gradually. Make it about THEM.
Instead of garbage going in and poison coming out — get the RIGHT inputs. Clean, continuous, real-time machine data. That's the carbon. AI turns it into diamonds: real insights, actionable guidance, decisions your team can trust.
Clean, continuous, real-time machine data. The heartbeat of every machine — captured through its electrical draw. Universal. Non-invasive. Air-gapped.
Actionable intelligence. Guided decisions. Patterns your team can trust. Expertise that scales across every shift, every plant.
Don't try to go from zero to autonomous. Build your AI strategy in stages. Be intentional about where you are on this progression. Stage-gate it. Don't skip steps.
AI helps your people do their jobs better. Remove the noisy stuff. Shift summaries, changeover guides, root cause suggestions. This is the easiest win — start here.
AI flags what needs attention before you ask. Not "what happened last week" but "what's happening NOW and what to do next." Your GPS, not your rearview mirror.
Rules-based agents handling triage. Clear guardrails: approve up to X, anything beyond triggers human intervention. Governance is key — define the escalation path in advance.
Every one of these was built on the same thing: the right problem, the right data, and the right people. When platforms are built right, the people closest to the problems become the solution architects.
J&J makes 80% of the world's sutures. Strong lean foundation, but machines that couldn't be connected through traditional methods. AI bridged old equipment to new intelligence — without touching PLCs, no cybersecurity risk. Team went from gut-feel to data-driven guidance across suture operations.
Full case study details coming soon.
A manufacturing professional in North Carolina hated uploading 800 work orders in Excel every other week. Using AI coding tools and a real-time data foundation, he built his own scheduling app — hundreds of orders processed in seconds. Then went further: built his own anomaly detection with custom alert thresholds. Five years ago, that's a six-figure consulting project. Today: one person, one weekend.
Customer managing production across multiple co-manufacturers — different companies, no shared visibility. Using AI coding tools and a secure API, they built a live production map across every co-manufacturer. Not because the coding was easy. Because the data foundation was already there.
A CRO needed to see which jobs were most labor-intensive, where to charge more, where the trends were going. By overlaying sales data on real-time production data, the team grounded pricing, capacity planning, and sales strategy in machine truth — not estimates.
Built a digital Plan-Do-Check-Act system for one of the largest steel manufacturers in the world. The CI cycle every lean practitioner knows — now powered by real-time data instead of whiteboards and sticky notes.
Everything in this playbook comes from working with 400+ manufacturers across five continents — from mid-market companies to Johnson & Johnson, Hershey, U.S. Steel, GE Aerospace, Coca-Cola FEMSA, and Kimberly-Clark.
Guidewheel — the first scalable way to bring AI to the factory floor
We found a way to read the “heartbeat” of every machine through its electrical signature — non-invasive, air-gapped, cybersecurity de-risked. 2.5 minutes to connect a machine. The oldest one connected: built in 1935. No open-heart surgery on your equipment. No PLC access. No OT network exposure.
That heartbeat becomes the data foundation for everything: real-time visibility, guided actions, AI-powered insights. Across our customer base: 1.4X average productivity improvement (third-party survey).
Simple beats complex. Now beats later. Start with one machine, one line, prove the value, then scale. Crawl, walk, run. The playbook in this session is universal — it works regardless of whether you use Guidewheel, another platform, or build your own. The principles are the same: right problem, right data, right people.
A $100M med device operation. Bottleneck lines at 55% utilization.
You move them to 65%. What happens?
Same machines. Same people. Same building.
The extra revenue drops almost straight to the bottom line — because your fixed costs don't move.
Regulatory alert: The QMSR took effect February 2, 2026 — replacing 21 CFR Part 820 and harmonizing with ISO 13485:2016. The FDA is now using AI tools (ELSA) to analyze complaint data, adverse events, and inspection history to prioritize inspections. A strong, real-time data foundation isn't just good practice — it's a regulatory advantage.
Four things you can do this week. Whether you're on the plant floor, in engineering, in quality, in the C-suite, or anywhere in between. No budget required. No IT ticket. No vendor call.
Operations:
"What are the top 5 causes of unplanned downtime on [your machine type]?"
Quality / Regulatory:
"Summarize the key differences between Part 820 and QMSR and what my team should prioritize first"
Engineering:
"What are common failure modes for [your component] and how would you design a test protocol?"
Leadership:
"Help me frame a business case for [your initiative] that connects to board-level priorities"
Compare what AI gives you to what your team says. The gap is your starting point. There aren't any experts yet. The barriers to learning have never been lower. What matters is what you choose to do with it.
• Know what data is going into which tools. Audit AI tool usage like you'd audit system access.
• Keep operational technology air-gapped. Do NOT connect AI directly to PLCs or your OT network.
• In med device: data integrity isn't just an efficiency question, it's a compliance question. Treat AI governance like your quality system.
Free or low-cost tools available right now. Start experimenting. The biggest predictor of success is whether the leaders use AI themselves.
Anthropic's AI assistant. Excellent for analysis, writing, coding, and manufacturing problem-solving. This website was built with Claude.
OpenAI's AI assistant. Strong general-purpose tool for research, analysis, and brainstorming.
Google's AI assistant. Good for research and integrates well with Google Workspace tools.
AI-powered code editor. Manufacturing professionals are using it to build custom apps on top of their data — no CS degree needed.
The factory down the road running the same machines, quoting the same jobs, serving the same customers — if they can see their floor in real time and you can't, they're not just faster. They're winning your next contract.
The window between early adopter and late majority is compressing faster than any tech shift in history. While you're planning, your competitors are deploying. The gap between “evaluating” and “behind” shrinks every month.
% of U.S. adults adopting each technology — years since mass-market launch
What took the internet a decade, AI did in three years.
ChatGPT reached 100 million users in 2 months. Instagram took 2.5 years. Nothing has ever moved this fast.
Sources: NBER / Federal Reserve (2024 & 2025), McKinsey State of AI (2025), Our World in Data, UBS, Deloitte
Carbon in. Diamonds out.
Problem → Strategy → People.
The foundation is the strategy.
“I can't control the wind or the waves, but I can learn to become an awesome surfer.”
AI advisor grounded in this session's frameworks
CEO & Co-Founder, Guidewheel
Lauren Dunford is the Co-Founder and CEO of Guidewheel, a World Economic Forum Technology Pioneer, TED speaker, and regular speaker at Davos and Stanford. She believes how we make things matters. A leading voice on “physical AI,” Lauren is bringing intelligence to the machines that power the real economy. Backed by top investors and scaling rapidly across five continents, Guidewheel already helps 400+ manufacturers—including Johnson & Johnson, Hershey, and Cargill—boost productivity by up to 2X while cutting energy costs and emissions by 15-20%.