AI in Manufacturing

When we hear “AI” today, we often think of LLMs (Large Language Models) like ChatGPT, Claude, Grok, and Gemini that are quickly becoming part of our daily lives. While AI broadly includes other models like decision trees, computer vision, and traditional machine learning (which are often what’s being referenced as “Industrial AI”), at Threaded we’re focused on opening up the value of LLMs in manufacturing. These models offer a unique ability to analyze factory operations, act as assistants and agents, and are super-approachable. To unlock their potential though, it’s important to remember that they are a tool that scales solutions, providing the right context in the right way is what enables them to shine, and you get the most gains when using them in a way that optimizes for AI + Human collaboration.

#AI is a Tool for Scale, not a Solution on it’s Own

From the first stone tool to modern automation, technology’s role has always been to scale solutions, not to invent them out of thin air. Manufacturing is no different, and the foundational methods of Industrial Engineering aren’t being replaced by AI. Rather, AI augments the work of modern engineers by harnessing the collective knowledge of thousands of books and experts encoded in an LLM. This means the future engineer is Human + AI, where both optimize for the same outcomes: efficiency, throughput, quality. AI excels at pattern recognition in complex systems and performing repetitive digital tasks, while humans bring situational awareness, physical-world judgment, and leadership. AI then becomes the amplifier that can help make your team smarter, faster, and remove the tedium. So engineers and builders can can focus on what we do best: solving meaningful problems and building great products.

#Context Matters

Ask an LLM “How do I optimize my factory?” and you’ll get a generic, (mostly) correct, but unfortunately not very actionable answer. This is because it lacks context - the details of your system, data, and objectives. Without that, it can’t offer very useful or trustworthy responses.

But how do you provide that context? LLMs can consider a limited amount of context at once, have no memory (you aren’t training the model while you interact with it), and your factory has a massive amount of relevant contextual data. Managing this context appropriately is essential to get value from an LLM, and there are a couple of ways to do it.

#Option 1: Training a Custom Model

You can train or tune a bespoke AI model on a curated dataset from your factory. This could be for any type of model (not just an LLM), and solves the context problem directly to enable solutions like closed-loop control or predictive maintenance for continuous processes (e.g., refining, chemical production, and some food/beverage manufacturing). The downside is that it requires significant resources: data engineering, research, model ops, and infrastructure to name a few. It also requires A LOT of data.

Training or fine-tuning a custom model is out of reach for most manufacturers, and doesn’t tap into the ever-expanding benefits of frontier models that you’re getting used to. You’ve probably noticed how fast AI is improving, and if you take this route you may find yourself spending more time and resources trying to keep up rather than actually improving manufacturing operations.

#Option 2: Managing Context for Frontier Models

Another option is to give pre-trained LLMs structured access to factory data. This allows them to navigate, interpret, and act without retraining. They have already been trained on a vast amount of public information, and approaching context this way allows us to leverage that training and unlock their full capabilities as intelligent assistants, analysts and and agents.

This is the approach Threaded takes, leveraging the power of frontier models through the familiar Value Stream framework. By allowing you to map your system and process, and then embed your operating data, we can combine frontier AI and your factory context to give world-class results with scalable, general-purpose AI tools for factory improvement. As we expand and enhance our platform and models improve, all users benefit immediately without retraining or infrastructure changes. Even better, as you add more data into Threaded, your insights and AI support continuously improve.

#AI + Human Collaboration

To collaborate, humans and AI must speak the same language, share context, and be able to work together effectively. For this to happen, the AI has to exist in an environment or tool where humans and AI can see the same system, use the same tools, and take the same actions. There shouldn’t be something that only AI can do, nor visa versa. This means AI can suggest actions, analyze data, and support decisions, but the human remains fully in control - accepting, rejecting, or modifying suggestions before they are implemented.

It doesn’t mean that we eliminate the need to go and see what’s happening on the factory floor, but instead AI can help level-up your team and get you to the right course of action, faster. This model guards against hallucinations, ensures accountability, and leverages AI where it shines: augmenting productivity rather than 1:1 replacement.

#Benefits of AI in Manufacturing

#1. Faster, Cheaper, Data-driven Analysis and Insights

Deming said it best:

“In God we trust. All others bring data.”

But analyzing, transforming, and understanding a mountain of data is tedious and difficult. This is made even more difficult for complex, interconnected systems like manufacturing. AI, however, excels at this. Using the value stream framework, it can synthesize data and generate actionable recommendations in seconds. When you combine that with your real-world experience in the system, it unlocks superpowers. No more weeks of thankless, tedious, off-and-on analysis. Instead, just ask “How can I double production?” and get a data-driven tactical plan instantly.

#2. Embedded Documentation and Systems Management

AI is already supercharging software engineers, because it is embedded in the tools where engineers write code and manage their codebase (using the same principles highlighted above). Similarly, AI embedded in Threaded helps industrial engineers map value streams, analyze systems, and create an manage their documents to unlock this same capability for physical world systems.

#3. A World-Class Assistant on Call

With access to your data, AI becomes a relentless co-pilot. Think of it as Jonah from The Goal (if you haven’t read it, you should!) available 24/7, who not only knows the theory but also sees your data and provides actionable insight—“Where are my top 3 bottlenecks?” or “What’s the best way to double production this quarter?”. Your AI assistant never gets tired, and is always down to help - no matter how crazy the idea or difficult the objective.

#How to Get Started

Start with what you know: your Value Stream. Capture your flow, process data, improvement ideas, and contextual documentation. Unlike Post-Its, AI can read and reason about this map - and suggest improvements.

As you expand, we can help you connect more systems: BOMs, training records, standard work, automation databases, ERP/MES platforms etc. The more data AI sees in the right time and context, the better insights it can provide. Threaded helps build this pipeline and integrates frontier AI into your workflow - securely and without training these models on your data.

With Threaded, AI becomes a tool not just for large enterprises, but to help every manufacturer scale. We manage the complex context in a platform set up for Human + AI Collaboration to help you scale your team, speed up problem-solving, and unleash productivity gains in manufacturing.