OpenBOM at Threaded at ACE 2026 by Aras: CAD File Agent, Product Memory, and AI for Engineering

Threaded at ACE 2026 by Aras: A Startup Space That Changed the Tone

Earlier this week, I went to Miami, FL to speak at Threaded, a free startup gathering space created inside Aras ACE 2026. Check the website and a great line of startup presented there – worth exploring.  The format was intentionally different from a traditional expo – a presentation room and an opportunity to meet with Aras management team and went to the conference to meet Aras customers and community. 

The event was organized by Michael Finocchiaro, known to most of the community as Fino, and sponsored by Aras. I want to give them both proper credit upfront, because what they built matters.

Bringing emerging companies into the middle of an enterprise PLM event, in a format that actually encourages conversation, is harder than it looks. Threaded made room for that. It created a setting where people could talk about real problems instead of performing for prospects. That is a meaningful difference, and it showed in the quality of the discussions throughout the event.

Why OpenBOM Was There

I have loved the Aras PLM community for years. For OpenBOM, Threaded was an important moment. In my session “From CAD Files to Digital Thread: Launching AI File Intelligence for Engineering,” I covered the transformation we see in the PLM industry – moving to the next level From System of Records to Digital Thread to Product Memory. I covered it in my Beyond PLM article PLM Single Source of Truth: Why It Was Never the Final Answer (And What Comes Next). 

The presentation was not only about a product launch. It was about a larger architectural argument for where engineering and manufacturing software needs to go next.

At the center of that argument was our announcement of CAD File Agent for SOLIDWORKS, which I presented as the first step toward the OpenBOM Product Memory Platform. I was also able to share the broader roadmap: Review Agents focused on BOM validation and change impact analysis, and Flow Agents designed to move product knowledge downstream into ERP, supplier coordination, and other systems. That roadmap context matters, because CAD File Agent only makes full sense when you understand what it is the beginning of.

Engineering Data Is Fragmented by Nature

The opening argument in the presentation was also the most fundamental one. Engineering and manufacturing data is fragmented by nature. That is not a technology failure. It is not a sign that companies are poorly managed. It is a structural condition of how products are actually built across teams, tools, suppliers, and functions.

The slide said it plainly: every handoff is a gap, spreadsheets and email are not failures of discipline, and no single system has a complete view of the product. Engineers work in CAD. Buyers work in ERP and supplier portals. Manufacturing teams operate their own systems. Service and customer signals arrive through still other channels. Between all of them, product knowledge moves through Excel files, local drives, email threads, PDFs, and conversations. That is not an edge case. That is how the work actually flows.

This is why I think discussions about the digital thread often become too abstract. In theory, everyone agrees that connected information is better. In practice, product knowledge breaks at handoffs, and the symptoms are familiar: wrong revision used in procurement, cost data missing from BOM, sourcing information separated from design context, wrong BOM sent to manufacturing. The handoff is exactly where knowledge gets lost.

Why One Platform Is Not the Answer

This leads to the second argument. For years, engineering software has been shaped by a compelling platform promise: consolidate more lifecycle activity in one environment, and the fragmentation problem goes away. The vision was not wrong. The problem is that real company workflows are simply broader and messier than any single platform can absorb.

To make this point vivid, I used a slide showing a Dassault Systèmes keynote presenting the 3DEXPERIENCE platform as the mainstream SOLIDWORKS environment. There is nothing wrong with 3DX vision, which is great but it is not complete and there are many things customers do beyond the platform. I listed what companies actually run day-to-day. Multiple CAD tools and PDM vaults. Different PLM platforms. Spreadsheets and email for BOM handoffs. Shared drives and local desktop files. Supplier portals and contract manufacturer packages. ERP systems with their own data structures. And manual coordination filling the gaps between all of them. That is not an exception. That is the default for many organizations I see.

The better question is not how to solve all problems using a single platform or system. It is how to build a connected layer that works across the systems companies already have. That question is what led me to Product Memory.

Product Data vs. Product Memory: What Gets Lost

One of the central slides in the presentation used an iceberg metaphor to draw a distinction most PLM and digital thread discussions still miss. Above the waterline is what systems capture reasonably well: part numbers, BOM structures, released drawings, CAD files, approved vendor lists, change orders, and ECOs. Below the waterline is something far more fragile and far more valuable.

What gets lost is the reasoning. Why a component was chosen. What trade-offs were discussed in the design review. Which supplier was rejected and why. How a change was justified. What a customer issue revealed. The thinking behind every decision. That knowledge disappears when projects move fast, when teams change, when people leave, and when work moves across organizational boundaries.

Systems are reasonably good at storing what was released. They are much weaker at preserving the context of how a decision was made. Yet that context is exactly what organizations need later, especially when products evolve, suppliers shift, new team members join, or AI systems begin operating on top of historical data.

This is the gap Product Memory is meant to address. It is not a slogan. It describes the missing contextual layer between isolated data records and real organizational knowledge.

The OpenBOM Product Memory Platform Vision

Once the problem is clear, the platform vision becomes straightforward to explain. The goal is not to replace existing systems. It is to connect the knowledge that currently lives between them.

In the presentation, I introduced the OpenBOM Product Memory Platform through a three-stage flywheel: Capture, Review, and Flow. Capture brings product knowledge into connected structure from CAD, files, and engineering work. Review validates and contextualizes what changed, why it changed, and what it means downstream. Flow moves structured and contextual product knowledge across teams, suppliers, ERP, and lifecycle stages.

This model avoids two common mistakes. The first is assuming that all enterprise knowledge can be cleanly centralized from day one. The second is treating AI as a thin conversational layer floating above disconnected systems. Product Memory requires both connected data and connected context. It must capture the traces of engineering work, make them reviewable, and make them usable downstream. That is what makes it a flywheel rather than a feature.

CAD File Agent for SOLIDWORKS: The First Step

This is where the announcement fits. CAD File Agent for SOLIDWORKS is the Capture stage of the Product Memory Platform, the point where engineering knowledge begins its journey into a connected system.

The capabilities it delivers are: CAD file capture with references and dependencies, structured design organization, multi-user coordination and access control, complete change history with revisions and comments, one-click restoration to any prior project revision, and agentic workflows with conversational product context. The important part is not the file storage itself. It is that engineering knowledge becomes connected and traceable from the moment design work happens.

It is also worth noting what this product is built on top of. OpenBOM has spent the last several years building the data foundation that makes this intelligence layer possible: a connected collaborative workspace used in production across engineering and manufacturing teams, native CAD and engineering data integration, and structured pathways from engineering BOM into procurement, inventory, ERP, and financial systems. That track record was recognized this year with a G2 Top 50 CAD/PLM Products 2026 designation. CAD File Agent is not a concept built on top of a slide. It is built on top of a real production platform.

In many organizations, file management is treated as a narrow administrative task. But file context is the beginning of product context. If you do not know how files relate, how they changed, which dependencies matter, and what design activity they represent, it is very hard to build a meaningful digital thread beyond them. CAD File Agent starts where most SOLIDWORKS teams actually are today and turns that starting point into something connected and useful.

What Comes Next: Review Agents and Flow Agents

Product Memory only becomes powerful when it moves beyond capture. The roadmap has three stages, and CAD File Agent is the first.

Coming next are Review Agents: focused on BOM review and validation, engineering change impact analysis, and missing data identification. After that come Flow Agents: ERP synchronization, supplier coordination, PDM/PLM sync, and CAD agents for platforms beyond SOLIDWORKS.

The sequence is deliberate. First, capture engineering knowledge in context. Then review it with intelligence that understands change, completeness, and downstream effects. Then make it flow into the rest of the business. Too many systems attempt to automate the final movement without first establishing the knowledge layer that makes that movement trustworthy. A BOM Review Agent is not a chatbot commenting on a spreadsheet. A Flow Agent is not a connector with a better interface. These agents act on connected product knowledge, and that is a fundamentally different proposition.

What Threaded Revealed About AI in Engineering

One of my takeaways from Threaded was that the conversation is maturing. The broader ACE agenda covered AI in manufacturing, digital thread orchestration, traceability, and product development workflows. That mix suggested something real: the question is no longer whether AI will appear in PLM, CAD, and manufacturing software. That part is settled. The deeper question is what knowledge layer those systems will actually run on.

If they run on fragmented records with little context, they will remain shallow. If they run on connected product memory, they can begin to do something meaningful. Threaded created space for that kind of architectural conversation, not just product pitches. In that sense, it felt less like a side event and more like an early signal about where the market is heading.

Conclusion and Final Thoughts

Thank you to Fino for organizing Threaded, and to Aras for sponsoring it and creating room for this kind of startup and architectural conversation inside ACE 2026. The format worked, and the discussions it enabled were worth the trip.

For OpenBOM, this was an important milestone. We announced CAD File Agent for SOLIDWORKS as the first step toward a broader Product Memory Platform that will also include BOM Review Agents and Flow Agents. The roadmap is just beginning. But the foundation is real, the vision is clear, and the timing feels right.

The product organization that builds a Product Memory layer today will compound that advantage every year. The knowledge you capture now is the institutional memory your AI systems will run on tomorrow.

Register for CAD File Agent rollout to get access. 

Best, Oleg