A Guide to Threaded's AI Assistant

Your manufacturing operations generate a massive amount of data every day. Hidden in that data are the key insights you need to increase efficiency, reduce waste, and solve complex problems. But how do you find them?

At Threaded, we’ve built an AI Assistant that acts as your manufacturing intelligence co-pilot, designed to help you optimize operations, identify bottlenecks, and make data-driven decisions - faster.

This guide explains how our AI works, so you can leverage it to its full potential.

#Table of Contents

What are LLMs?
The Foundation: Training Data and Manufacturing-Specific Knowledge
Context: The Key to Relevant AI Insights
From Conversation to Execution
Maximizing Your Impact: Key Takeaways
The Future of Manufacturing with AI

#What are LLMs?

Large Language Models (LLMs) are advanced AI systems that have been trained on massive amounts of text data (training data), including much of the information publicly available on the internet. This training allows them to understand, interpret, and generate human language.

Once deployed, LLMs can receive and process inputs rapidly and effectively, and communicate meaningful insights in clear, plain English.

In the context of manufacturing, LLMs excel at:

Analyzing operational data and identifying trends across multiple metrics
Translating complex data into actionable insights for plant managers and engineers
Understanding industry terminology relevant to manufacturing processes and workflows
Generating detailed reports that combine quantitative analysis with qualitative recommendations

Threaded’s AI Assistant leverages models from frontier AI providers including OpenAI, Anthropic, and Google. By using frontier AI models, Threaded ensures you’re getting the most advanced AI capabilities available.

#The Foundation: Training Data and Manufacturing-Specific Knowledge

But where do these capabilities come from? The answer lies in the training data that gives the AI its deep manufacturing knowledge.

#What is Training Data?

Training data refers to the collection of documents and content that AI models learn from during their development. The frontier AI models used by Threaded are trained on the enormous set of information that is publicly available on the internet - including web content, books, articles and specialized datasets across a broad range of topics.

The includes extensive manufacturing content such as:

Lean manufacturing and Six Sigma methodologies
Engineering textbooks and manuals
Industry publications and technical standards
Equipment manuals and maintenance procedures

As a result, the AI can “speak manufacturing” fluently and understands manufacturing terminology and concepts without requiring explanation.

When you mention “Overall Equipment Effectiveness (OEE),” the AI already understands the complex relationships between availability, performance, and quality.

#Training Data ≠ Your Data

Unlike internet algorithms, AI models are not trained on your private data. Models are trained and validated by frontier AI labs before public availability, and further validated by Threaded for use in the platform before inclusion.

Your data is used at runtime-analyzed temporarily to generate responses -and is not provided to the AI labs to train or improve the underlying models.

#Context: The Key to Relevant AI Insights

Training gives an LLM its foundational knowledge, but delivering relevant, useful answers requires something more: context. Context refers to all the information the model has access to while generating a response—including your specific question, recent interactions, and relevant system data.

#Context Engineering

“Context Engineering is the discipline of designing and building dynamic systems that provide the right information and tools, in the right format, at the right time, to give a LLM everything it needs to accomplish a task.”

— Phil Schmid, Senior AI Relation Engineer at Google DeepMind

At Threaded, we apply advanced context engineering practices to ensure our AI Assistant always has the information it needs—structured properly and delivered at the right moment—to provide meaningful, accurate, and task-specific answers.

#The “Context Window”: An AI’s Working Memory

LLMs operate within a technical constraint known as the context window—a limit on how much information can be processed at once. Think of it like working memory: just as humans can only hold so much information in their minds at once, AI systems have limits to how much data they can consider simultaneously.

Manufacturing organizations generate extensive operational data. Think of every shift report, quality control data point, equipment history, and training record. Sending everything at once would be both technically infeasible and cost-prohibitive.

In Threaded, core system and process information is sent with every request and specialized tools dynamically pull in detailed information and data when needed. This keeps response times fast, maintains relevance, and ensures your AI assistant always operates within its technical limits—without compromising the quality of its insights.

#Automatically Included Organizational Context

When you interact with Threaded’s AI Assistant, it automatically includes context about your operation based on the information you’ve connected and defined in the platform.

This includes a high-level summary of:

Value stream maps showing your specific process flow and operational metrics
Equipment and tooling data including maintenance schedules and performance history
Training matrices tracking your operators’ skills and certifications
Bills of materials detailing your product components and specifications
Shift reports capturing live operational data and production metrics
Action items and improvement initiatives showing your ongoing optimization efforts

The more robust your organizational setup in Threaded, the deeper and more relevant the AI’s insights.

#User-Added Context

Users can guide the AI further and add specific context by:

Detailed written descriptions of specific challenges, goals, and constraints in the chat
Attaching external files to the chat including documents, spreadsheets, images, and equipment manuals
Specifying value streams to focus analysis on particular manufacturing processes using the @ button
Attaching Previous conversation summaries to maintain continuity on complex projects using the @ button
Referencing specific people, parts, processes, tools and actions using in-line @mentions

#From Conversation to Execution

Now that you understand how context works, let’s look at how Threaded structures AI interactions—from conversation to execution.

#How Conversations Work

Every new conversation with the AI Assistant begins with a clean slate—the AI does not remember previous conversations as a default, but builds knowledge within a session as the discussion progresses and context is added.

Best Practices for Conversations:

Keep conversations focused: Address a single goal or a set of closely related questions within one conversation.
Start new conversations for new tasks: When your focus shifts to a new challenge or topic, it’s best to start a new conversation.
Use summaries for continuity: To expand on a previous analysis, include a summary of the prior conversation as context in the new one.
Attach relevant files early: Provide all necessary documents, data, and context at the beginning of a conversation for the best results.

#Tools: How the AI Gets Things Done

The true power of Threaded’s Assistant comes from its ability to act, not just analyze. Tools enable the AI to pull data, run structured analysis, and take meaningful steps within your workflows.

Threaded’s AI Assistant has access to three distinct categories of tools:

#Data Gathering Tools

Part of the Threaded’s context engineering system, these tools let the AI retrieve detailed data when needed, while always having access to a high-level organization snapshot.

For example, the AI can see that you have 50 shift reports completed this month, but only retrieves the specific contents of those reports when responding to a user query that requires considering and analyzing this information. This targeted approach optimizes system performance and operates efficiently within AI model constraints.

#Structured Analysis Tools

Threaded AI includes specialized analytical frameworks that bring proven manufacturing methodologies directly into the assistants analysis. These tools ensure that AI analysis follows consistent, proven methodologies rather than ad-hoc approaches.

Key analytical frameworks include:

OEE Analysis: Comprehensive evaluation with component breakdown, value stream impact assessment, and quantified improvement recommendations
Action Item Review: Summarizes and prioritizes tasks based on operation impact and resource needs
Value Stream Analysis: Detailed process flow examination with bottleneck identification, capacity analysis and suggestions for optimization

#Execution Tools for Agentic Workflows

These tools turn analysis into action and allow the AI assistant to act as a digital agent.

For example:

After identifying an issue impacting OEE, the Assistant proposes a corrective action plan
It suggests action items to implement the plan- including details about assignees, due dates, and financial impact.
You review and approve the plan—and the action items are automatically created and assigned.

This human-in-the-loop approach ensures that valuable insights drive real operational improvements while you maintain full control.

The agentic capabilities of the AI assistant are continuously improving with both Threaded features and model enhancements. Regardless of the action it takes, it will always do so on your behalf and with human approval.

#Maximizing Your Impact: Key Takeaways

To get the most out of the AI Assistant, keep these principles in mind:

Input and Maintain Rich Data in Threaded: The more comprehensive and accurate data you maintain in Threaded (value streams, shift reports, maintenance logs), the more insightful and specific the AI’s recommendations will be.
Give Specific Context: Start your conversations by clearly stating your goal and attaching any relevant documents, spreadsheets, or images. The more focused your input, the better the output.
Start a New Conversation When Changing Focus: Keep AI analysis focused on a specific goal. To build on previous work, provide a summary of a prior conversation as context.
Leverage AI-Powered Actions: Don’t just ask for analysis. Use the AI to create action items, assign tasks, and build project plans to turn insights into real-world improvements.

#The Data Advantage: Better Data = Better Insights

Even with a small amount of data, Threaded’s AI Assistant can deliver valuable insights by analyzing patterns, identifying obvious bottlenecks, and helping you structure your improvement efforts.

As you add more system context and data over time, the AI’s recommendations become increasingly precise, connected, and impactful.

For example, connecting equipment history, operator training, and quality control data enables the AI to identify the root causes of recurring issues—like linking downtime to specific machines, shifts, or parts—and automatically suggest targeted improvements. The more complete your data, the more valuable and actionable the AI’s insights become.

#The Future of Manufacturing with AI

Threaded’s AI Assistant represents a new paradigm where artificial intelligence amplifies human expertise. By combining pattern recognition capabilities with deep manufacturing knowledge, contextual awareness, and practical tools, we’ve created an AI system that truly understands your manufacturing challenges and helps solve them.

The result is operational excellence through AI that reduces time-to-insight from weeks to seconds, identifies opportunities that might be missed by manual analysis, and adapts continuously with your operations to provide increasingly valuable insights.

As manufacturing becomes increasingly complex and competitive, leveraging AI will be essential for maintaining operational excellence and driving competitive advantage.

To dive deeper into Threaded’s philosophy and approach to AI in manufacturing — read our blog post on the topic: AI in Manufacturing.