Think AI Podcast - WHEN MACHINES START TALKING - AI IN MANUFACTURING | EP. 3

Episode Date: March 11, 2026

🎙️ Inside the Factory: Real-Time AI Intelligence for Manufacturing | Think AI Podcast Ep. 3Your machines already know what's wrong — but nobody's listening. Dave Goyal takes you inside 4 real manufacturing companies to show how real-time data and AI transformed their operations. Plus: a 4-layer framework to get started and 7 AI use cases you can implement now.⏱️ In this episode:00:21 – Why your factory machines are smarter than your spreadsheets03:15 – The data silo problem: ERP, MES, and PLCs that don't talk06:30 – 98% exploring AI, only 20% ready — here's why08:45 – Case Study 1: Professional audio equipment13:00 – Case Study 2: Global outdoor adventure gear17:30 – Case Study 3: Medical devices (FDA compliance & traceability)22:30 – Case Study 4: Fall protection safety equipment27:00 – The 4-layer framework: ERP → Shop Floor → Analytics → AI32:00 – 7 AI use cases: predictive maintenance, computer vision QA & more42:00 – "My ERP handles everything" — why that's not enough46:00 – AI Tip of the Day: Analyze any spreadsheet in 60 seconds50:00 – Series recap and what's nextIf this changed how you think about manufacturing data, subscribe and share it with a plant manager, CIO, or operations leader who needs to hear it.🔗 Links & Resources📖 Real-Time Business Intelligence Mastery — DaveGoyal.com🌐 DaveGoyal.com💼 LinkedIn: Dave Goyal🎙️ Ep 1: Personal Journey & First AI Assistant | Ep 2: Healthcare & AIManufacturingAI #RealTimeData #SmartFactory #DataAnalytics #ThinkAIPodcast

A machine on your factory floor knows something right now. It's running 3% slower than yesterday. It knows a bearing out. It knows the batch it's just finished had a slight temperature drift. But no one is listening. Because the data from that machine is sitting in a PLC or in MES systems or log files, nobody knows. Meanwhile, the production manager is looking at a spreadsheet from last week, making decisions

with old data about problems that has already happened. That's how most manufacturers operate today and it's costing them millions. Welcome to the Think AI podcast. Each week we talk about the most exciting AI research, tools, case studies and more. I'm your host Dave Goyer and I have a lot. I've been working behind the scene in data and AI for over 30 years. Whether you are an AI expert, skeptic, or something in between, this podcast is for you. I have spent years inside manufacturing companies, discrete manufacturing, medical device manufacturers,

companies that make professional audio equipment, outdoor adventure gear, safety equipment. I've seen their ERP systems, their MES systems, their MES systems, their dashboards and more often their lack of dashboards and I have helped them to do something most manufacturers still haven't done I have helped them make their data talk let's call it talking reports in real time now last week's data not yesterday data not right now last week data is what they generally see we are talking real-time data. That's what this episode is all about.

Welcome back to Think AI podcast. I'm Dave Goyer. Today we are going inside manufacturing real stories from real factories and real-time intelligence. And they use cases on AI, those are going to separate the survivors from the ones who don't make it.

AI has to be there. By now, you know that. If you run a manufacturing company or lead operations, IT, or supply chain in one, this episode is for you. If you're in healthcare, if you're any other industry, stay with me. Because the principles of real-time data, NAI applies everywhere. Let's get into it. So let's start with an honest statement.

Most manufacturers are data-rich and inside poor. They have data everywhere. their ERP system has it, machine generates it, quality inspection produce it, their supply chain creates it, but it's scattered, it's siloed, and by the time anyone looks at it, it's too late and it's too late to act. Let me paint the picture for you, a typical mid-sized manufacturer, let's say they make 500 or 5,000 units a day, and they have an ERP system, but maybe it's a It's a well-known system like Oracle, SAP, NetSuite or something else that ERP handles orders, inventory, purchasing, financials. Some of them also have a manufacturing execution system called AMES that tracks what happened on the shop floor, which machines are running, what's being produced, how fast.

A few have PLCs, programmable logic controllers on their machines. These are the brains of the equipment. They control the motors, the sensors, the actuators. Here's the problem. The ERP knows what was ordered. MES knows what's being made. PLC knows what's the machine and how the machine is running. But they don't talk to each other.

The ERP does not know the machine is running slow. The MES does not know the orders just changed. The PLC does not know if there's a material shortage coming. Now, each system is smarts on its own, but together they are strangers to each other, living in the same house. That's what we call silos. And here's what's the cost that you will incur. When your systems don't talk, you get delays.

Delays in production, delays in shipping, delays in invoicing, and delays in decision-making. You get waste over production because nobody knew what the order and how the order has changed, because the quality issues wasn't caught until after the batch was done. Excess inventory because the purchasing did not know what the floor already had. You get downtime. A machine fails. Nobody saw it coming. And the line stops, you lose hours. Sometimes days. A recent industry survey found that 98% of manufacturers are exploring AI. But guess what?

only 20% feel prepared to use it. Why? Because their data is not ready, not an AI problem. Their systems are disconnected. Nobody has given them a clear path from where they are to where they need to be. That's the gap we fill. That's what Think AI does. And that's what I wrote about in my book called Real Time Business Intelligence Mastery. I wrote that book specifically for manufacturing leaders,

CIOs, CTOs, VPF IT and operations, because I kept having the same conversation over and over again. Smart leaders, good companies, sitting on mountain of data they could not use in real time. The book lays out the framework, how to connect your system, how to build real-time dashports, how to go from reactive decision-making to proactive intelligence

and to build data culture and governance. Everything I'm about to share with you in this episode comes from that same framework, tested and proven, in real factories with real results. All right, let me take you inside four manufacturing companies we work with. Different industries, different products, same core challenge. None of them had real-time visibility into their operations. All of them have transformed how they have transformed how they are. run their business using data. You will see that. Now, the first company makes professional

audio equipment, high-end speakers, amplifiers, mixing consoles, product use in concerts, studios, corporate AV systems around the world. And this is something I love, sound engineering. This is discrete manufacturing. Every unit is assembled from hundreds of companies, circuit boards, enclosers, wiring, firmware, every one or each one tested it before it ships. They had an ERP system. It tracked order and inventory, but the production floor was a black box. Management could tell you how many units shipped last month, but they could not tell you right now how many units are in progress. Where each unit is in the assembly process or not, which workstation has a bottleneck.

And they were running a sophisticated manufacturing operation with a rear view mirror. So we connected their ERP to a real-time analytics layer. We pool production data from the floor, test readers, let me repeat. So we connected their ERP to a real-time analytics layer. We pulled production data from the floor, test results, assembly completion stages, and quality checks. We built dashboards that showed live the status of every work order, every production line and every quality matrix. What's the outcome? The production manager could see at 10 a.m., the line 3 was running 15% behind, not the end of the day, not in the weekly

report. Right now, they could see which components are running low before they run out. The purchasing team got alerts, not when the shelf was empty, but when it was trending towards empty. Lead times dropped, inventory accuracy improved, and the leadership team finally had a single source of truth. You'll hear me saying that a lot. Typically for their operations, they need to have single version of truth. That's what real-time intelligence does. Let me say it again. That's what real-time intelligence does. It doesn't just show you what happened. It shows you what's happening and what is about to happen. The second company makes outdoor adventure gear, sleeping bags, camping mats, dry bags, technical outdoor equipment sold globally. They had a unique challenge. Their products were made across multiple facilities in different countries,

components source from different suppliers, finally or final assembly in different locations. Their ERP had the order data, but the supply chain was a mess of spreadsheet, emails, and phone calls. Nobody had a single unified way of materials where the materials were, where production stood across facilities, whether they could fulfill a large retail order on time or not. And in outdoor gear, timing matters. Miss the season and you miss the sale. Retail buyers do not wait. We build a global operations dashboard, data pool from the ERP across facilities,

unified into one single view. Inventory levels, production status by facility, order fulfillment, tracking, supplier delivery timelines, all of that. For the first time, the operation team could see every could see everything, all facilities, all products, all in one place. They spotted supply chain delays weeks before it came to production stoppages. They re-balance the entire production across facilities when one site was overloaded and another

had capacity. They reduced excess inventory by matching production more tightly to actual demand. The CIO told me something I'll never forget. He said, Dave, we used to make decisions based on gut feeling and experience. Now we have decisions based on facts and we are faster than we have ever been. Now, that's the shift from gut to data, from delayed to real time, from hoping to knowing. The third company is the one that hits close to my home for me, a medical device manufacturer. making orthopedic supports, braces, rehabilitation devices, I wear braces, it's close to my heart.

If you have listened to Episode 2, you know healthcare is in my DNA. The medical device manufacturing sits right at the intersection of healthcare and manufacturing. Everything I've learned in both worlds comes together when you get into medical device manufacturing. Now, medical device manufacturing is not like making consumer products. Every single device is regulated. You have to meet the FDA requirements, other compliance requirements. Quality has to be documented at every step. Traceability is not optional.

It is the law. If a brace fails on a patient's knee, you need to trace that device back to the actual batch, the exact materials, the exact production run, the exact operator. And that's the level of data discipline required. This company had the discipline, that's not the issue, but the data was trapped in disconnected systems. Quality records in one database, production logs in another, ERP in third, compliance let's say just in paper binders in excess. Finding the answer to a simple question, what material went into batch number, let's say 4721, could take even hours. So we build a unified data

platform, ERP data, production data, quality data, compliance data, all connected, all queryable, all ready to analyze, all in real time. We build traceability dashboards, enter a lot number, see every material, every production step, every quality test, every operator in seconds and not ours. So what we build? We build a unified data platform, ERP data, production data, quality data, companies, and compliance data all connected, all queryable, all in real time. We build traceability dashboards.

So enter a lot number, see every material. Every production step, every quality step, every operator, and not in seconds but in hours. So it's actually reversed. Not in hours but in seconds. We build quality trend analytics. Instead of catching defects after they happened, the system showed trends. A slight drift in the measurement, a pattern in a particular material lot, a correlation between specific supplier and higher rejection rates. Now, that's shift from catching defects to predicting them, is the difference between quality control and quality intelligence.

Audit preparation time dropped dramatically. What used to take weeks of pooling records become days. and more importantly, production quality improved because the team could see issues forming an act before they became defects, before they reached patients or their customers. When you are making medical devices, good enough data is not good enough.

It has to be absolutely right. It has to be real time and has to be traceable. That's our specialty, discrete manufacturing. medical device manufacturing where the data accuracy isn't nice to have its patient safety requirements. Fourth company makes fall protection safety equipment, things like harnesses, lanyards, self-reacting lifelines, equipment that construction workers fully utilize or the industrial workers wear on day-to-day basis. I want you to think about that for a second. If this product fails, someone falls, someone lose the life, someone may get hurt and even dies, like I said. That's the weight behind every data in this company.

They had an ERP system running their orders and financials, but production planning was largely manual. They relied on tribal knowledge, experienced planners who knew from years of data, doing it, how to schedule production, manage materials, and hit delivery dates. And that worked until it did not. So when those experienced planners retired or left or found a better job, the knowledge walked out, the door with them. New people came in, but they didn't have 20 years of intuition, knowledge, experience and wisdom.

They may have experience, but not the knowledge of that company. They needed data. We connected their ERP to a production intelligence platform, real-time visibility, material availability, production progress, and shipping timelines. We built planning dashboards that took the guesswork out of scheduling based on the current inventories, open orders and production capacity. The system recommended optimal production sequences. We've also built compliance and testing analytics.

me say that again, we've also built compliance and testing analytics. Every harness gets tested, every lifelines gets inspected. The data was now tracked, trended and visible in real time. On time delivery improved, material waste decreased and the new team members could make decisions with data, not just instinct. But here's what mattered most. The quality data gave This company confidence. Confidence that harness leaving their facility met specs. Every lanyard was tested. Every lifeline was safe. Because when your product keeps people alive, confidence isn't a business matrix. It is a moral obligation. All right. Four industries, four stories, same pattern. Let me now give you the framework because if you are a manufacturer

listening to this you might be thinking that's great there but where do I start? Here's what or here's where let me stop. Here's where you start. AI driven data analytics in manufacturing has four layers. Think of it as building a house. You can't put the roof on before the foundation. So let's talk about each layer. Layer 1. Connect your ERP. Almost every manufacturer has an ERP.

That's your foundation, your orders, your inventory, your purchasing, your financials. But most companies use their ERP as a record keeping system. They put the data in. They pool reports out of it. That's it. The first step in turning your ERP from a record keeper into a real-time data source, that means connecting it to a modern analytics platform, pulling data out automatically, not once a day, not once a week, continuously.

If you do nothing else from this episode, do this, connect your ERP to a live dashboard, see your orders, your inventory, your production status in real, time, that single step changes how you make decisions. Let's talk about layer 2. Connect your shop floor. If you have an MES system, connect it. If you have PLCs on your machines, pull that data. If you have manual tracking, barcode, scans, operator inputs, quality checks, digitize it and

connect it. The goal is simple. Know what's happening on the floor right now. What happened last night or in the last shift? Which machines are running? At what speed? What's the output versus what's the target? What's the bottleneck?

When your floor shop data connects to your ERP data, something powerful happens. You can see the full picture. Your order came in, your materials were available, your production started, and here's where it is right now. when it will ship, that insight is out of the world. Trust me. End-to-end visibility, that is layer two. Layer three. Build your analytics. Now, that's your data is connected. And when it is connected, you can start asking smart questions. Examples. Where's my real cost data per unit, including downtime, scrape and rework. Which productions or which

products are most profitable when I factor in actual production time, which supplier deliver on time, which ones consistently delays. What's my true on-time delivery rate? Not the one which we report, but the real one. Where am I losing capacity? Is it machine down time, change over time, waiting for materials? These questions sound basic, but I promise you, most mid-sized manufacturers cannot answer them in real time and some can't even answer them. When you can, you make better decisions faster with less waste. Layer 4, which is the crux of everything here, add AI. This is where it gets exciting. This is where the future is heading. Once your data is connected, clean and flowing in real time, AI can sit on top of it, and

and do things humans simply can't. And that brings me to the AI use cases I want you to walk you through. Here are the AI use cases that matter most for manufacturers right now, not five years from now, right now. And I'm sharing these because I want you to see what's possible, whether you are already in this journey

or getting started. Let's talk about these use cases. Use case one, predictive maintenance. Every manufacturer has equipment that breaks down. When it does, the line stops. You lose production, you lose money, you scramble for parts. Traditional manufacturing and maintenance is either reactive, fix it when it breaks it or preventive.

Maintain it on a schedule when it's needed or if not. AI offers a third option, predictive, and sometimes you can extend it to prescriptive, a talk for some other episode. Sensors on your equipment track vibration, temperature, pressure, speed. AI model learns what normal looks

like when your pattern shifts. When something starts to drift, the system flags it. Not after the machine breaks it, but before, which is really useful and valuable. You schedule the repair during planned downtime.

You order the part before you need it. You avoid the emergency. Companies using predictive or prescriptive maintenance are reducing and even fixing unplanned downtime by 30 to 50%. That's not a projection. That's what is happening right now. Use case 2, AI powered quality inspection. In discrete manufacturing, quality inspection is often manual. A person looks at a part, checks dimensions, check for defects, and then signs off. It's really slow. It's subjective.

And human eyes gets tired. You make mistakes. Computer vision, AI that sees, can now inspect parts at production speed. It catches defects that human misses, scratches, cracks, dimensional variations, color inconsistencies, and it does it consistently. The thousand inspection is as accurate as the first one. No fatigue. For medical device manufacturers, especially, this is critical.

FDA inspects quality, we all know it. Consistency documentation is the key. AI powered inspection delivers all three. Now use case three, demand forecasting. How much should you produce next month? Next quarter. If you guess wrong, you either overproduce or tie up your cash in inventory or overpresent. entry or overproduce and miss sales.

AI can analyze your historical sales data, your seasonal patterns, customer ordering, trends, market signals, and it gives you forecast that's significantly more accurate than the manual planning. Plug that forecast into your ERP. You have some advanced planning systems but they are no good. Your purchasing team knows what to order. Your production team knows what to schedule. warehouse teams what's coming APS systems are pretty strict and you can flex it

using AI based systems that's not just planning that's synchronized planning driven by data not by gut use case for intelligent production scheduling most manufacturers schedule production manually or with basic ERP tools don't account for real-world constraints what if your scheduling system could consider current machine availability, operator skills, materials availability, their training, order priority, change over times and due dates all at once in real time. That's what AI scheduling does. It evolves thousands of these scenarios in seconds, sometimes in milliseconds based on the power of the system, but it finds the optimal sequence. It adjusts dynamically when things

change. One customer told me we went from spending two hours a day on scheduling to just 15 minutes and the schedule is even better. Use case 5. Supply chain risk detection. Your supply chain is only as strong as its weakest link. A delayed shipment from one supplier can cascade across entire production plan. AI can monitor your supplier performance, their delivery times, quality scores, lead time trends, and it can flag risk before they become problems. Supplier X has been shipping two days late for the last three orders. AI catches that. Alerts your purchasing team, you find a backup or you adjust your production plan.

This is actually prescriptive analytics. Proactive, not reactive, that's the pattern. Use case 6. Energy optimization. Manufacturing uses a lot of energy and energy costs are rising. AI can track energy consumption by machine, by line, by shift. It identifies patterns. Which machines are running inefficiently?

Let me say it again. Which machines are running inefficiently? When is energy consumption highest? Where can you reduce without affecting output? One manufacturer found that a single production line was consuming 22% more energy than the night shift. Because of a calibration issue and that nobody noticed. AI just caught it, they fixed it, saved tens and thousands a year. Use case 7.

This one is more advanced, where it will be useful where the manufacturing is heading fast. A digital twin is a virtual replica of your factory, your machine, your production lines, your processes, all modeled digitally. You can simulate changes before you making them. What happens if I add a second shift? What if I arrange or rearrange this production line? What if I switch to a different supplier for this component? You test in the digital world, you deploy in the real world, less risk? faster decisions.

7 AI use cases, all practical, all available today. You don't need to do all 7 at once. Start with 1. This one will solve your biggest pain point. So select that one. For most manufacturers, that's predictive maintenance or real-time production visibility. Start there. See the value.

Then expand. That's how we work. One layer at a time, one win at a time. Until your entire operations run on real-time intelligence. It's a journey, not a milestone. Now, I know what some of you are thinking. Dave, my ERP handles everything. I have spent so much money on it.

We have been running on it for 10, 15, 20 years. Why do I need AI? Fair question. Let me answer it directly. So your ERP is essential. I'm not telling you to replace it. I would never say that. Your ERP is the backbone of your business.

It tracks your orders, inventories, your finances, and this is all critical. But your ERP was designed to record transactions, not to predict outcomes, not to analyze patterns, not to give you real-time intelligence. Some of them are adding it, but not entirely as that is not their focus. So think of it this way. Your ERP is like a good filing cabinet. Everything is organized, everything is labeled. and everything is stored.

But a filing cabinet doesn't tap you on your shoulder and say, hey, you're about to run out of aluminum stock in just three days. And by the way, your biggest customer just increased their orders by 20%. That's what real-time intelligence does. It sits on top of your ERP. It doesn't replicate or replace. It makes it smarter. And here's the good news.

You don't need to rip out your ERP. You don't need to buy a new system, you connect what you already have to an analytics layer. You know, tools like Snowflake, Databricks, Microsoft Fabric, to name a few, but you can choose anything. You start pulling insights from that data that's already there. Most of the companies I work with, they didn't need new data. They needed to see that data that's already in their systems, in real time, in context, with the right length. And for skeptics who say AI is overhyped, I hear you. A lot of AI marketing is overhyped.

But what I'm talking about isn't hype. It's connecting your ERP to a dashboard. So your production manager does not have to open six different spreadsheets every morning. It's alerting you purchasing team. Let me stop. It's alerting your purchasing team before a start. stock out, not after. It's showing your quality team a trend before it becomes a defect. That's not hype. That's just good business. So no, your ERP is not enough, but your ERP plus

intelligence, that's powerful. And that's exactly what we need you to build. All right, time for our AI tip of the day. In episode one, I showed you how to build your First, AI Assistant. There are more advanced techniques, but I showed you something really basic. In episode 2, I showed you how to summarize any document in two minutes. Today's tip, use AI to analyze your own data instantly. Here's what I want you to do. Step 1.

Take any spreadsheet you use regularly. A production report, a sales report, an inventory list. Step 2. Open any AI tool. supports file uploads like GPT, Claude, Copilot. I like Claude Co-work but that's on your desktop, a little more expensive. That's a good assistant for my analysis. Step 3, upload the file, then type this prompt.

Analyze this data and tell me the top five patterns or trends you see. Highlight anything that looks unusual or concerning. Give me three recommendations to improve. based on what you find, use plain simple language, talk to your AI. That's it. In 60 seconds, AI will analyze data that would take you for an hour to process manually. And here's what amazing. The AI does not just give you numbers.

It gives you insights. It tells you stories hiding inside your data spreadsheet. So data to insight, to action. Examples? Your scrape rate has been increasing. 2% per week for the last month. Here's which product line is driving it. AI will tell you.

Your top 3 customers account for 62% of revenue. Here's how that concentration creates risk. AI will tell you. You are on time or your on time delivery rate dropped in last three weeks. Here's the correlation that supplier delays from vendor A. AI will tell you. These are not hypothetical examples. This is what AI does.

does when you give data. Now for AI curious, this is the fastest way you see AI create real business value. Upload one spreadsheet, see what happens. For the enthusiast, start doing this weekly every Monday, upload your key reports, get an AI-powered briefing before your first meeting. This all can be automated. For the AI skeptics, compare the AI analysis with your own and see if that catches anything you missed. I bet it will. That's your tip. One spreadsheet, 60 seconds, try it. So three episodes, three worlds. Episode one, I shared my personal journey from a kid in India who could not walk to building one of the top data and AI consulting firms in the country three times recognized as INC 5,000 awardee. In episode two,

I took you inside healthcare, how we build claim systems, data strategies and patient intelligence platforms for non-profit hospitals. In six AI, let me say it again, and six AI use cases that are transforming health care right now. In episode 3, manufacturing, which is this one, four factories, four real stories, and real-time intelligence. Seven use cases and a framework you can start implementing tomorrow. And here's the through line across three episodes. Here's the commonality across three episodes. Data is not just the technology problem. It's the leadership decision. It's the strategy issue. Every organization

I've worked with had the data, every single one. The data was there inside their ERP, inside spreadsheets, inside their EHR, inside machines. What they did not have was a strategy to connect it, the framework to make it real-time. And a partner, such as us, who understood both the technology and the business. That's what Think AI does, and that's why I wrote Real-Time Business Intelligence Mastery to give you the playbook. Now, this might be the end of our first three episodes, but it's not the real-time business. end of this podcast. I've got a lot more coming, deep dives into specific AI technologies,

conversations with data and AI leaders, who you think are doing incredible things, live walkthrough of AI tools you can use today, and I'll keep bringing you the AI tip of the day because the best way to learn AI is to use AI every single day. If these three episodes help you, if they change how you think about data, about AI, about what's possible for your business, here's what I would love you to do. Subscribe, share this podcast with anyone who needs it, a CEO, a operational leader, a plant manager, someone who's sitting on data, they don't know how to use it or any enthusiast or even a skeptic.

And drop me a message. Tell me what you want to hear about next. tell me your data challenges, tell me your AI questions, I read every single one and I mean it. You can find me on LinkedIn, Dave Goel or at Dave Goel.com. If you find the framework I've been talking about, interesting, grab the book Real Time Business Intelligence Mastery from Davegoel.com. It's the playbook for manufacturers. For any leader who knows data is the answer, but doesn't. know where to start. It's a great book. I'm Dave Goyer and this is the Think Yeahi podcast.

Thank you for being here all three episodes. It means more than you know. I'll see you in the next one. You have been listening to Think Yeah, Podcast with Dave. Take one idea from this episode and turn it into action.