Energy Efficiency with IO-Link and AI – Practical Insights from achtBytes

Hello, friends of IoT, and welcome to a new episode of the IoT Use Case Podcast. Today we’re talking about a topic many from the production world will recognize: energy and compressed air monitoring, CO₂tracking, and power monitoring. We’ll look at how to make the best use of this data, what to consider during implementation, and how AI can be applied to real-time data. For that, I’ve invited today’s guest, Alwin Jung, IoT Consultant at achtBytes. You’ll hear more in a moment about what exactly they do. As always, take all the insights and best practices you can for your own projects. I wish you lots of fun with this episode. You can find all the information as always at iotusecase.com or in the show notes. So, let’s get started.

Welcome, Alwin — great to have you here.

Alwin

Hello, Madeleine. Thank you for the invitation, I’m happy to be here.

Very nice, I’m excited too. We’ve known each other for quite a while, and today we’ll be talking about your projects and a few best practices. That should be exciting. How are you doing these days? What’s new at your end? How was your week?

Alwin

Things are very busy for us right now, especially in August when quite a few people are on vacation. Over the weekend I was at the Junior Bowl, the American football youth finals. It took place here in Schwäbisch Hall, hosted by the Schwäbisch Hall Unicorns. Unfortunately, not with a great outcome for our team, but it was an exciting day nonetheless. Maybe next year will be better.

Hmm, okay. Do you have a personal connection to that, or did you just go to watch the game?

Alwin

Yes, I have a personal connection. I used to play for the Schwäbisch Hall Unicorns in the youth team for five years — as a defensive tackle and offensive guard back then.

Very cool, then greetings to your region. You’re based in Schwäbisch Hall, right?

Alwin

Exactly. I’m working from home today because we’re doing some remodeling at our headquarters, but we’re generally based in Schwäbisch Hall.

Okay, so your headquarters is there too. Very nice. Greetings to everyone in the region who’s tuning in. Maybe to start with you personally: you’re an IoT Consultant at achtBytes. I know you mainly from the projects you implement with customers — practical IoT solutions around energy and compressed air monitoring and more. You studied business informatics, right? To start off, I wanted to ask: what excites you most about IoT projects? Not everyone works in this field. Was there a particular reason you got into it, or how did it come about?

Alwin

I find IoT projects incredibly exciting, maybe also because of my background in football. American football fans love statistics — everything is measured there. It’s quite similar in the IoT world, just in a different context. It’s about collecting data, analyzing it, and generating value for further development. That’s what fascinates me. Also, every company is different, with its own requirements and standards you need to adapt to. That makes the field so diverse.

Exciting. Maybe a quick introduction to achtBytes for those who don’t know you yet. You’re what one might call a corporate start-up of the STEGO Group. We can talk more about that in a moment. You develop cloud-based IoT platforms that connect various sensors from production environments. We’ll get to your specific use cases shortly, but I know you have a strong focus on IO-Link as a standard. Can you briefly explain the background of the STEGO Group? How are you connected, and how did it all come about?

Alwin

STEGO Elektrotechnik is our parent company. We’re part of the group and were founded as an additional pillar to strengthen the software business. STEGO originally comes from control cabinet climate management — that’s where we are absolute experts. At some point, the idea came up to expand into software. The initial idea was to understand how our sensors perform under real conditions, not just in lab tests. From that project, achtBytes emerged as a corporate start-up — with a clear focus beyond temperature monitoring, addressing topics like energy monitoring, retrofit, and particularly compressed air monitoring.

And who are your typical customers? Do you focus on certain market segments or company sizes?

Alwin

Our main focus is on the industrial mid-market, the classic hidden champions. Energy-intensive industries benefit the most from our solutions. For them, energy consumption is a key factor and a major cost driver.

So, for example, steel production or plastics manufacturing?

Alwin

Exactly — steel, plastics, but also food production. Those kinds of industries.

Okay, I assume you can’t share customer names. I always like to ask for references, but that’s probably tricky.

Alwin

Unfortunately, no.

All right. Then let’s still talk about your customers’ use cases without going into too much detail. Can you name some typical use cases you implement together with them? I mentioned a few earlier, but which ones are you currently focusing on?

Alwin

Typical use cases include power monitoring, CO₂ monitoring, and compressed air monitoring. The main goal is to gain deeper insight into energy consumption and create transparency to make well-founded decisions for the future. Another focus is retrofit projects, meaning upgrading existing systems with modern sensors to capture the necessary data in the first place. One example: many customers only have a single sensor at the main power line that measures overall energy consumption for billing purposes. Everything beyond that remains a black box. They don’t really know where the energy is actually used. Their only reference point is often the machine’s nameplate, which only shows the maximum power consumption. That has little to do with reality when trying to calculate how much energy an individual production order actually costs.

Okay, you mentioned that you also help customers plan and determine how much energy a process or production order truly consumes. What are some typical questions or starting points that lead your customers to launch such projects? Where do you dig deeper into the data, and how do you approach that?

Alwin

A common example is comparing two machines to evaluate their performance. Right now, we’re seeing a clear trend toward replacing hydraulic machines with fully electric ones. Manufacturers often promise major efficiency gains, but only by measuring real data can you verify what those promises look like in practice. This allows us to see the true benefit of switching technologies. Based on that, companies can decide whether to replace additional machines or continue operating existing ones that are still performing well.

So it’s not just about comparisons, but also about strategic decisions, such as investments in new equipment. That’s interesting.

Alwin

Exactly. The goal is to build a reliable data foundation — moving away from estimates or gut decisions toward truly data-driven decision-making about the future of production.

Do you have an example from a customer — without naming names — that shows what such a project setup looks like? I imagine there are various sensors involved. You mentioned the main power line earlier. I’m not sure if those are STEGO sensors or others, and there are likely several machines. Can you describe what a typical setup looks like when you visit a customer and start a project? How does it usually work on site?

Alwin

We first look at what’s already in place at the customer’s site. In most cases, there’s already a power meter at the main line, for example from Janitza or a similar manufacturer. These devices can be integrated into our system. Typically, a STEGO sensor is then added — a current sensor that’s installed on the individual phases of each machine. These sensors are connected via a master device. The same master can also connect to other sensors such as temperature or compressed air sensors to provide a complete picture of energy flows. The master is linked to the company network, and together with the customer we set up a virtual machine that handles connectivity. It manages communication between the sensors on the shop floor and our cloud application.

Before we get to your cloud application and how it works technically, let’s stay for a moment with data acquisition. What are some of the typical challenges you encounter when connecting devices? You work with IO-Link, among others. Are there specific standards that are particularly relevant here, or common technical issues you run into repeatedly in customer projects?

Alwin

One typical technical challenge with the IO-Link standard is the cable length limit of 20 meters. That means you need multiple masters positioned throughout the facility and must plan carefully how cables are routed. Twenty meters doesn’t sound like much at first, but in a large production hall, that distance is quickly reached—especially if you don’t want to run cables across the entire building. That’s one of the biggest challenges. Aside from that, IO-Link allows for very easy integration thanks to IODDs and the plug-and-play connection of smart sensors. This keeps the configuration effort low, so data connectivity to our platform can usually be implemented quite efficiently.

Okay, interesting. I’m also curious to hear from those of you listening—what challenges do you face in practice? Feel free to share them in the comments or on LinkedI. I’ll also link your contact details in the show notes, Alwin.

[11:22] Challenges, potentials and status quo – This is what the use case looks like in practice

Are there any best practices or common pitfalls to avoid?

Alwin

My clear recommendation is always to choose the larger master version. There are usually models with four or eight ports, and we often see customers start with the smaller one to save costs. But once the first data is in the cloud and analysis becomes possible, new ideas quickly come up: “Can we also add a distance sensor?” or “Can we capture this additional value?” If you don’t have enough ports, adding more later becomes cumbersome. For a small additional cost, the larger version gives you double the number of ports and much more flexibility in the long run. Ideally, you equip the production hall with masters at several points right from the start. This way, you can easily add new sensors later when needed. From then on, it’s really just: attach the sensor and get started.

You mentioned it’s smart to go with the larger version from the beginning since each machine usually has multiple sensors or may need more in the future, saving you the hassle of upgrading later. Did I understand that correctly?

Alwin

Exactly.

And besides that, are there any other best practices?

Alwin

I also recommend planning a switch or network connection close to each master. The masters can be connected directly via network cable, but the special connectors required are quite expensive compared to standard CAT7 cables. So, it’s worth planning the network layout carefully to keep cable runs short and reduce costs. A well-thought-out plan for where to place the masters not only simplifies integration but also saves money in the long run.

Okay, so in practice it’s about finding the right balance: do I use IO-Link masters, switches, or a combination of both? Is that an either-or decision, or how should we understand it?

Alwin

It’s really about placing the masters in areas that already have some level of network connectivity. That way, you don’t have to run long network cables across the entire building. You create central points where the masters are installed, and additional central points where these masters are interconnected. The result is essentially a two-tier network structure that remains both efficient and easy to manage.

[14:38] Solutions, offerings and services – A look at the technologies used

And in the end, the goal is to gain exactly these kinds of insights — for example, to compare two machines, determine whether replacing a machine makes sense, or analyze data such as: what was my pressure level yesterday? Analyzing this data is the next step after acquisition. So, where do the data go, and how does that process work on the software side?

Alwin

At the customer’s site, we have a virtual machine — a VM — that handles data transmission to our cloud. From there, the data are mirrored, and the assets are represented in the cloud so they can be fully visualized and analyzed.

By VM, you mean a virtual machine environment, right?

Alwin

Exactly, a virtual machine. From there, the information is transferred into our network. In the cloud, I can see exactly which sensors are active, what they measure, and I can even adjust parameters via the platform, since the sensors connect via plug and play. Once the data are collected, different analyses can be performed on the software side. For example, in power monitoring, the three phases can be combined into a total value, or you can calculate downtimes and similar KPIs based on that. The data are then clearly visualized in dashboards.
We’ve now integrated an AI assistant into our system called MIA. The idea behind it is that MIA acts like a colleague within the company — so you can simply say, “Just ask Mia.” She supports users in creating dashboards. The trend of the past few years is clear — AI is becoming increasingly important, both in industry and everyday life. You just have a text field where you type what you want to know or create, and the AI delivers the right answer or visualization. That’s exactly how we’ve implemented it.

Cool. So what kind of things could you ask MIA? Have you already tested some initial prompts — typical inputs you’ve tried out?

Alwin

Exactly. MIA isn’t designed to generate new data or calculate things that don’t exist. Her focus is clearly on visualization. You can ask her things like: “MIA, show me Machine 1 compared to Machine 2 based on last month’s production quantities,” or “Show me the performance over the past three months,” or “How has our CO₂ consumption changed in the last two months?” You can also say, “Overlay the compressed air values for me.” MIA then decides on her own which type of visualization is best suited for that request.
The advantage is that through visual representation, you immediately see if something doesn’t look right. AI systems sometimes tend to “hallucinate” data, similar to what’s known as the Wikipedia effect. But in the case of dashboarding, that’s not a real issue because you can spot instantly if, for example, the wrong time period was selected or the chart doesn’t make sense. That’s why MIA can be trusted in a business context — she’s, so to speak, not a very good liar.

I see. And does MIA access live IoT data directly, or does she work with a database you set up together with the customer?

Alwin

Exactly. Our dashboards can display both live and historical data, depending on the selected time period. If I want to see data from the previous month, the system automatically pulls the historical values. MIA herself doesn’t handle how the data enter the system — that’s already taken care of. We reliably receive both live and past data in our platform. MIA’s sole responsibility is deciding how to visualize the data to solve the user’s problem. So if the manager asks, “Can you show me how our CO₂ footprint has improved over the last three months?”, all it takes is a single request to MIA, and she automatically creates the right view. There’s no need to manually configure charts or dashboards.

I see. So when you talk about AI dashboarding, you mean automating the creation of visualizations through AI, allowing users to make individual queries for their specific use cases.

Alwin

Exactly. Normally, you’d have three or four fixed dashboards that you’ve built manually in advance. When a new question arises, you’d have to add extra parameters or set up new columns in Excel to combine values. With MIA, all of that goes away. You simply ask the question, and MIA provides the data in the best possible visualization.

Very cool. So far, we’ve talked a lot about process parameters — data from the OT world, like compressed air monitoring. Do you also access IT system data, for example from tools or enterprise systems, to answer specific questions? Or are you currently focusing mainly on data from the automation layer?

Alwin

We work primarily with data from the automation world. Our goal is to make existing sensor data visible and transform it into actionable insights.

Very nice. Do you have an example from a completed project that shows what a solution looks like in practice? What does the customer actually receive at the end when working with you?

Alwin

A typical starting point is compressed air monitoring. With just a few sensors, you can already gain a lot of valuable insights. Once the monitoring is running and data are being collected, customers quickly develop a sense of how their system actually operates. For example, they can see when the compressor really needs to be switched on in the morning, when it can be turned off in the evening, and whether the current pressure level is even necessary. They can also see whether the entire compressed air system is fully utilized and how to avoid unnecessary energy losses, such as from weekend operation.
Another example is power monitoring. Here, customers can see exactly how much energy a specific machine or production order consumes. This also allows them to calculate and share the CO₂ footprint of a product or order.
By using IO-Link, we remain highly flexible. The system can be expanded step by step. Through the master structure, new sensors or data sources can be added at any time, allowing the project to grow gradually — always aligned with the customer’s current questions and needs.

Maybe one last question on that. Not all products are IO-Link ready or use that standard. How do you handle it when a customer doesn’t have IO-Link-capable devices? Do you only support IO-Link, or are there other integration options?

Alwin

Our main focus is on IO-Link, and of course, it’s ideal when customers already use IO-Link in their environment. However, we’re also working on integrating other systems and enabling direct communication. For example, in power monitoring, where devices from manufacturers like Janitza are often already installed, we can integrate them through existing interfaces. For that, we just need information from the customer side about which smart interface can be connected. Such integrations can also be implemented individually on a project basis.

If you’re listening and have a similar use case — or would like to exchange experiences on how to implement such projects — feel free to reach out to Alwin. You’ll find his LinkedIn profile and contact details in the show notes, along with more information about achtBytes.

[23:59] Transferability, scaling and next steps – Here’s how you can use this use case

What can we look forward to in the future? You already mentioned the topic of AI dashboarding. Are there any other developments you’re focusing on?

Alwin

Yes, we want to give MIA even more capabilities, especially around configuration. The goal is to simplify processes such as alerting. Instead of manually setting an alarm limit, you’ll simply be able to tell MIA, “Please set up an alert for this case.” We’re also planning to introduce what we call artificial KPIs — derived sensor data calculated from existing values. Beyond that, we want to make system integration for customers even easier. The initial setup is currently quite demanding because many systems have to be connected. Our goal is to reduce that effort and help customers get to actionable data faster.

Do you currently have any initiatives where you’re looking for pilot customers to co-develop these innovations with you? Or are you collaborating with partners on this?

Alwin

Yes, we’re currently looking for pilot customers for a new technology layer. The idea is to bypass the entire IO-Link stack and communicate directly with the sensor.
Bye! The goal is that you simply connect the sensor to power, place it where needed, and the tracking starts immediately. For this project, we’re actively looking for pilot customers and expect to be ready in Q4 2025.

So, Q4 2025. We’re recording today on September 11, and this episode will likely air in September or October — so if you’re listening now, you’re already closer to that rollout. Very nice. Thank you so much for this insightful and practical conversation — especially for the clear explanation of the use case around compressed air monitoring, energy optimization, and the resulting insights. I also found it fascinating to hear your best practices from past projects and how your offering is structured. Thanks again from my side. And I’ll give you the final word, Alwin.

Alwin

Thank you very much for having me. If anyone has questions or wants to exchange ideas, feel free to connect with me on LinkedIn. See you soon.

Maybe we’ll do an update episode on your AI topic at some point. Thanks again, and have a great rest of your week. Take care, bye.

Alwin

Bye.