A conversation between Ayush Tiwari, who leads ThingWorx Product Management at Velotic, and Adam Gąsiorek, CTO and co-owner of Transition Technologies PSC, recorded at Hannover Messe 2026. The two discuss what manufacturers are actually asking for when it comes to AI, how a semiconductor customer uncovered hidden quality issues in data everyone had written off as noise, and what it takes to build a working manufacturing co-pilot in a matter of days.

The setting: a roundtable at Hannover Messe 2026

The conversation below took place at Hannover Messe 2026, during a closed roundtable session we co-hosted with Velotic Software: “Your AI wishlist when the data does not fight back.” This time TT PSC joined not just as a participant, but as co-host. Big thanks to Velotic Software for the invitation, and to Ayush Tiwari for making it happen.

The session brought leading manufacturers together to talk candidly about where AI actually helps on the shop floor, and where it stalls. TT PSC shared two concrete examples:

  • MFG Copilot, built on ThingWorx® from Velotic™, an agent-based interface for understanding and managing operations in real time.
  • Predictive Quality, combining ThingWorx® AIoT, industrial data streams and AI to detect early quality signals hidden inside what operators often treat as “warm-up noise”, long before issues become visible at SPC (Statistical Process Control) or customer level.

The interview that follows unpacks both: the predictive-quality case in depth, and the thinking behind the manufacturing co-pilot. A short demo of the co-pilot is at the end.

Inside the demo: MFG Copilot in action

The use case Adam describes above, shown live.

To show manufacturers what this looks like in practice, we walked through a short demo of the MFG Copilot, an AI agent supporting real-time decision-making on the shop floor.

It begins with the simplest possible question: “What’s going on in production?” From there, the agent surfaces risks, flags issues, and lets you drill down into a specific line. Behind that single answer it pulls together data from several sources at once, including machine telemetry, events, and operational systems, and turns them into one clear, actionable picture.

In the scenario, the agent detects a problem at a welding station. It analyses machine-level data, identifies the root causes, overheating and excessive vibration, and assesses the impact on both production and materials. Finally, it generates recommended actions and assigns responsibilities, so the right person can act faster and with better context.

That is the shift worth noting: this is not data visualisation. It is an interactive, AI-driven interface for understanding and managing an industrial environment. The difference between reading about a problem and being guided through it.

Key takeaways

  • AI does not change the business goals. Manufacturers at the roundtable were still chasing uptime, quality, and worker productivity. AI is the enablement layer, not the objective.
  • The most valuable data is often the data nobody is looking at. In the semiconductor case, unmodelled signals from an early process phase held the root cause of quality failures the entire time.
  • A working manufacturing co-pilot can be built in days, not months. The ThingWorx AI orchestration platform gives partners like TT PSC the building blocks to move fast.
  • Natural language changes who can act on data. When operators can query a production line the way they would ask a colleague, the gap between data and decision shrinks.

What this means for you

The semiconductor case in this interview is not an exception. Most manufacturers are sitting on process data they have never fully interrogated. The tools to do it now exist, and the barrier is rarely technical. If you are dealing with recurring quality issues, unexplained yield loss, or a dashboard that nobody acts on, the starting point is a conversation about your data, not a multi-year transformation programme.

Your factory data already holds the answer. Let's find it.
We turn industrial data into decisions. Let's talk about your use case.
Start the conversation

TT PSC is a Velotic Authorised Partner and certified Kepware system integrator, delivering industrial connectivity and AI solutions across Europe. The partnership combines Velotic’s ThingWorx and Kepware technology with TT PSC’s engineering expertise to help manufacturers turn operational data into measurable outcomes.

Interview transcript: AI on the factory floor

Ayush Tiwari

My name is Ayush Tiwari. I lead ThingWorx Product Management. Joining with me is Adam. Adam, do you want to introduce yourself?

Adam Gąsiorek

Sure. Welcome. My name is Adam Gąsiorek. I’m CTO and co-owner of Transition Technologies PSC. We help our customers digitise their processes and solve business problems with new technology, like augmented reality, artificial intelligence, and cloud computing.

Ayush Tiwari

Awesome. So we recently hosted a roundtable discussion with leading manufacturers to understand their AI use cases and the challenges that they have in implementing those. It was a great session, and we would love to share some of the details.

Adam is one of our leading partners who implement the technology that Velotic provides, specifically ThingWorx and Kepware, in different customer solutions. So Adam, why don’t you talk about some of the AI use cases we heard and the demonstration that you gave about one of the top use cases, Predictive Quality?

Adam Gąsiorek

Sure. Sure. First of all, thank you for the opportunity to be the co-host of the session. For me, it was great fun and a great experience to be able to sit down together with manufacturers, ask them what are the challenges, how they see the importance of data in solving the manufacturing problems, and also share ideas from our side, how we can take a pragmatic approach to optimise processes and generate positive impact at site scale, manufacturing line scale, and also on worker scale.

Ayush Tiwari

That’s really critical. And as we heard loud and clear, AI is kind of the enablement technology, but the business goals are still the same. Our customers were looking for ensuring high uptime of their assets, ensuring high quality, save cost, ensuring worker productivity. Worker productivity was one of the top use cases. So why don’t you tell us a little bit about the predictive quality use case that you’ve built using ThingWorx AI capabilities.

So ThingWorx, we launched this AI capability, which is an AI orchestration platform, and our customers and partners like yourself are building AI agents on top of the platform and solving specific customer use cases. So tell us a little bit about the predictive quality use case that you demonstrated to some of our manufacturers who joined this session.

Adam Gąsiorek

Absolutely. That’s a perfect example of how data perceived initially as a chaotic set of bits and bytes can be leveraged to find out the root causes of possible low quality in production batches.

This is one of my favourite examples of applying Artificial Intelligence, because it shows perfectly that questioning the status quo can lead you to finding benefits, to finding results, to solving the problems.

Let me give a brief description of what the problem was.

Ayush Tiwari

Yes, please.

Adam Gąsiorek

So our customer comes from the semiconductor industry, and they were facing a problem of claims and returns of their products, even though they have very complex, robust product quality checks in clean rooms. But still, there were some undiscovered opportunities of improving the process. The data seemed to be not structured or modelled well enough to find out the root causes.

We took a slightly alternative approach. We proposed to have a look at the unmodelled signals from the initial phase of the processes, something that was perceived as noise, as chaos. And we found out, based on historical data analysis and based on applying various techniques of machine learning, that we can find with 100% accuracy the root causes: specific process parameters in the filtering of chemical substances in that undiscovered phase of the process. Something that was initially perceived as unusable data.

Ayush Tiwari

Awesome. The implementation of the AI capabilities that ThingWorx offers is enormous. It was great to see the manufacturing co-pilot that you built in a matter of two days, using those AI capabilities, and we were ready with the demonstration for the customers to show them the art of the possible. We want to talk a little bit about the manufacturing co-pilot.

Adam Gąsiorek

Sure, sure. That’s a great example of how we can see AI as a multiplier. It is a multiplier of efficiency and productivity, right? With this co-pilot approach we move from simple visualisation and dashboards to the next level. So you can talk with your data. You can use your natural language. You can ask multiple queries that normally would take time.

And of course, machine-readable data allows that. But those co-pilots, those agent-based interactions, give you the feeling as if you are talking with a manufacturing supervisor who advises you on what may be the status of the production line, which machine is underperforming, what may be the root cause or root causes, what may be the next action.

So that brings data more into information, actionable information, so that immediately you can react. Acting on data, right, is the purpose of having AI capabilities in ThingWorx.

Ayush Tiwari

Absolutely. That’s awesome.

.