English
Français
Delivering scalable agentic intelligence to your digital thread
Earlier this week, Aras introduced InnovatorEdge AI as one of an expanded set of services built on the Aras InnovatorEdge framework. Edge AI allows Aras customers to scale their engineering AI program by leveraging the rich context of their digital thread while providing a practical governance model ensuring continued compliance, traceability, and protection of intellectual property.
We are excited to introduce the first set of agents and agentic services built on Edge AI that deliver immediate value to any Aras Innovator® customer (leveraging the fact that Edge cloud services can be integrated with any version of Aras Innovator, whether SaaS or hosted on private infrastructure). But they also serve as building blocks for other agents and, in the spirit of Aras’ commitment to adaptability, can be configured to meet the specific needs of our community across the many markets and product lifecycle Aras Innovator is used.
Edge AI and Edge AI agents will be central topics at ACE 2026 as we explore the theme of Adaptive Intelligence and our vision for the future of PLM shaped by an intelligent digital thread, AI-native user experience, and a disruptive agility to respond to change. You can register for the event here.
These first Edge AI agents fall into two categories: 1) Conversational Agents that are available through the Aras AI Assistant console embedded in Innovator; and 2) Task Assistants that are invoked by a user or when a specific condition is met to execute discrete workflows or analyses. The figure below compares these agent interaction patterns with other approaches that will also become important in an AI-native PLM environment.
Discover new insights from your digital thread
The new conversational agents integrated with the Aras AI Assistant are focused on Discovery (aligning to the Discover-Enrich-Amplify framework we use to segment PLM agentic services).
Analytics and Insights Agent
The Analytics and Insights service expands the capability of the Aras AI Assistant by enabling engineers and PLM users to ask analytical questions (i.e., questions with discrete quantitative or Boolean answers) in plain language, using voice or text, and quickly retrieve the product data they need without learning Aras Markup Language (AML), SQL, or advanced search syntax. The Analytics and Insights agent leverages InnovatorEdge AI Natural Language Analytics (NLA) service to translate user intent into precise queries, then returns results in a form that supports decision-making. When analytics are included, it can also aggregate and analyze the returned results, creating charts, trends, and summary statistics to help users interpret what they are seeing. In this mode, the agent does not rely on a separate knowledge base; it helps users ask better questions and consume answers faster.
Knowledge Analysis Agent
The Knowledge Analysis service, also embedded in the Aras AI Assistant conversational interface, leverages a unique implementation of GraphRAG (look for future articles and ACE sessions on Aras ThreadRAG). It builds on an advanced semantic search capability to analyze Innovator managed knowledgebases, including documents, figures, and other unstructured/loosely-structured data stores to find digital information related to the user’s prompt. It is capable of identifying relationships that may be implicit rather than explicitly modeled in PLM structures, for example, identifying connections between a document and many referenced part numbers, even when those relationships were never formally created. This enables richer conversations across the digital thread because users can ask questions that require meaning, linkage, and context, not just retrieval. The Knowledge Analysis service was built with the understanding that content changes over time. Exploration of relationships must be done in the context of when (e.g., what were the requirements that drove design decisions for the version of the product that just failed in the field – when that product was designed 10 years ago).
Enrich the context and connections contained in your digital thread
The new task-specific agents are focused on enriching your digital thread by inferring or extracting new structured data from thread-connected digital information and establishing “relationships of record” that can be leveraged in the future. These task-assistant agentic services are invoked by users or can be embedded in broader workflows. In practice, they are supervised or semi-supervised services where the agent suggests new relationships that are approved by human users for codification within the digital thread.
Requirements Ingestion Agent
The Requirements Ingestion agent ingests source documents in PDF, Excel, and Word formats, identifies requirements within them, and converts the raw text into structured requirements in Aras Requirements Engineering. The key outcome is standardization; it separates content from presentation and enforces consistent requirement types so teams can operate on requirements as governed objects rather than formatted text blocks. This shifts human effort away from tedious transcription and toward validation and intent, where domain expertise matters. It also supports similarity analysis across imports, enabling teams to compare newly ingested requirements with those already in the system to identify duplicates and confirm what is genuinely new.
Smart Variant Agent
The Smart Variant agent analyzes large populations of BOMs and produces a variant model that includes features, options, and rules, and can generate configured BOMs as output. It can also incorporate supporting documents such as catalogs and order sheets to learn the vernacular for how products are sold and configured, which is often not fully inferable from the BOM structure alone. This matters because the business meaning of configuration, what defines racing versus off-road, often lives in documentation rather than in part lists. The outcome is a practical on-ramp from engineered-to-order sprawl to configure-to-order discipline, grounded in the product structures and the language the business already uses.
BOM Assistant Agent
The BOM Assistant addresses the reality that critical product structure information often arrives in spreadsheets, PDFs, and other formats from many sources, especially in supplier collaboration scenarios. The agent infers how incoming data maps to the correct PLM structure, interpreting hierarchical indentation patterns, level indicators, and other conventions to normalize the data into a standard structure inside Aras Innovator. This reduces manual mapping and rework when the input format was never designed for clean import. It also sets the stage for lifecycle workflows such as understanding what changed between a Rev A BOM and a supplier-provided Rev B BOM, so teams can apply deltas intentionally rather than repeatedly re-importing entire structures.
Scaling AI in PLM needs to stay pragmatic
Each of these agents targets a persistent friction point in enterprise PLM, getting answers from product data faster, discovering meaning and implicit relationships across the digital thread, turning unstructured requirements into governed objects, accelerating the shift from engineered to order to configure to order, and normalizing BOM inputs that rarely arrive in perfect form.
Just as importantly, they are not point solutions embedded into the application logic of a piece of PLM software. The InnovatorEdge / InnovatorEdge AI framework ensures they can be deployed, governed, monitored, and improved as operational capabilities inside real product development environment evolves. Beyond clever automation, natural language understanding, integration of generative AI, and application of powerful data science, the real value of agentic AI in PLM will hinge on repeatability, trust, and the ability to scale across teams, programs, suppliers, and lifecycle stages. PLM leaders will need to work with digital transformation and AI teams to harness the power of AI without creating a parallel universe of disconnected decision-making that cannot scale.
This first set of agents address very real PLM use cases. But they are intended to seed the market with concepts. As we expand the InnovatorEdge AI ecosystem, the goal is to give customers and partners a governed foundation for building and adopting agents that act on the digital thread with the same discipline enterprises already expect from PLM.
