A Major Leap Forward in Industrial Information Processing, Answer Accuracy, and Agentic Intelligence
InSkill, the leading platform for industrial AI copilots, today announced the release of its third generation Agentic AI Engine. This release revolutionizes how industrial organizations capture, curate, and deploy AI copilots & expertise for complex equipment.
With this release, InSkill introduces an upgraded in-copilot expertise capture framework, enhanced agent capabilities, and a major overhaul of its prompting and AI infrastructure powered by the latest Azure and OpenAI APIs. Version 3 enables the platform to deliver faster, more reliable, and better-formatted responses.
“This is our most important release yet because it ensures the highest quality answers, the ability to extend LLM capabilities with custom agents, all while providing industry leading safeguards against hallucinations”
Key Advancements InSkill’s v3 Agentic AI Engine
Capturing Worker Expertise Automatically
Industrial workers are retiring at unprecedented rates. InSkill’s Version 3 Agentic AI Engine empowers organizations to capture worker expertise directly in the field from the same tool they already get their answers from.
Experts are prompted to provide their expertise. The AI Engine then leads them through the process of capturing new problems to solve along with the steps to resolve it, the safety requirements, parts, tools, and necessary worker skills. All solutions are saved as unpublished and can be curated, modified, reviewed, and finally published by backend personnel.
“We’ve built powerful new expertise capture tools directly into the copilot chat. It is now easier than ever to capture field expertise or create private copilot extensions with site-specific procedures,” said Jim Hansen, CTO at InSkill. “By default, new solutions are saved in an unpublished state, allowing review and curation before they go live. This safeguards data quality and prevents premature publishing of user-generated content.”
Industrial sites can also utilize this feature to capture site-specific expertise or processes. This expertise is private to the site and is not visible outside the organization, even if the copilot was originally provided by the machine’s manufacturer.
Asset Notes and Contextual Intelligence
Users can also use the copilot to attach notes and insights directly to individual assets. This keeps teams up to date and enables copilots to deliver even more precise, context-aware recommendations. This granular linking between expertise and equipment transforms every machine into a living, continuously improving knowledge source.
InSkill copilots are unique in that the information sources powering answers are always ensured to be the right sources for the equipment and the worker who is using it. InSkill copilots understand the subtle nuances between machines in industrial environments and always deliver answers that are context aware. This means that answers are safe for both workers and machines. Safety protocols are always enforced to prevent injury or damage to equipment or parts. Information is always delivered based on worker skill level.
Create Custom Tools and Agents
InSkill Agentic AI Engine V3 introduces a next-generation agentic architecture built for tool creation and custom AI behaviors. Users can design and deploy purpose-built agents that automate workflows, perform diagnostics, or assist with maintenance tasks built using the InSkill platform.
The upgraded APIs significantly enhance tool recognition and agent performance, resulting in smoother execution and more predictable outcomes.
The version 3 release of the InSkill agentic AI Engine offers even more granular control and extensibility, while ensuring accuracy and remaining free from hallucinations. It retains all the benefits of agentic responses while fine-tuned for industrial applications to eliminate hallucinations and prioritize safety for both workers and equipment.
Prompting Overhaul for Precision and Presentation
InSkill’s entire prompting pipeline has been rebuilt from the ground up. Leveraging the latest Azure OpenAI endpoints, the system now:
- Returns cleaner, better-formatted responses with numbered steps and actionable guidance including rich text and even mathematical equations or calculations
- Supports custom prompting for different machines and copilots
The result is a more natural, polished, and useful interaction for operators.
The InSkill team is continuously optimizing prompting to address the rigors of industrial environments.
Detailed Follow Up Questions
Version 3 of InSkill’s Agentic AI Engine takes follow-up questions to the next level. Each time a worker receives an answer, the copilot now automatically suggests relevant follow-up paths: “Would you like to learn more about A, B, or C?” This allows users to explore deeper layers of knowledge with a single keystroke. This makes complex topics easier to navigate and transforms every chat into an interactive learning experience. Whether troubleshooting, training, or verifying a procedure, InSkill copilots now guide users through a structured, conversational journey that accelerates understanding and captures context along the way.
3rd Generation of the InSkill Agentic AI Engine: Built for the Agentic Future
The third iteration of the InSkill Agentic AI Engine is engineered to handle agents more effectively, manage context across follow-on interactions, and enable site-specific copilots that extend OEM-provided capabilities.
With these advances, industrial sites and OEMs can now:
- Capture worker expertise in the copilot
- Add custom procedures and safety workflows
- Extend OEM copilots with private local knowledge directly from the copilot chat
- Maintain full privacy and data control within their site
- Create and deploy custom tools and agents
- Enhance context awareness by capturing asset information
- Expand operator learning with conversation follow up questions
This marks a key step toward truly agentic industrial AI.
All InSkill copilots don’t just assist, but learn, capture worker expertise, and adapt in real time (with curation and approval).