The challenge
Behind every product a manufacturer designs today sits the accumulated weight of everything it has designed before. Materials change. Customer specifications shift. Order books fragment into smaller, more customized runs. Yet the underlying engineering challenge stays the same: how to keep drawing on decades of accumulated know-how without repeating the same design work, project after project. As product portfolios grow more complex and customization becomes the norm rather than the exception, the ability to systematically reuse what a company already knows — rather than reinvent it — is quietly becoming one of the clearest differences between manufacturers that scale efficiently and those that don’t.
This challenge is especially acute in Engineer-to-Order and Make-to-Order manufacturing, where companies can process dozens of new CAD projects every week. In the aluminum extrusion die industry, a design that already exists — in a slightly different form, filed away years earlier — often solves the problem a customer is asking about today. Finding it, however, is harder than it sounds. Keyword search and manual archiving cannot capture geometric similarity, and traditional image-comparison methods become computationally impractical once a database reaches thousands of profiles. The result is a familiar and costly pattern: design teams quietly re-engineer components that already exist, losing time, consistency, and the compounding value of the company’s own engineering history.
One of the companies tackling this challenge is Phoenix International S.p.A., one of the leading international players in aluminum extrusion dies. With over 50 years of experience, 10 manufacturing plants, and one of the largest engineering teams in the sector, Phoenix has built an exceptional design archive and a depth of engineering know-how that made it a natural industrial partner for Intellico.
50+
years of experience
10
manufacturing plants
>95%
simple profiles
80%
complex profiles
The approach
More than a technology implementation, this story represents the outcome of a strategic partnership that combined Phoenix’s unique industrial know-how, engineering excellence, and one of the sector’s richest design archives with Intellico’s cutting-edge expertise in AI-driven image similarity and geometric intelligence. The value created emerges precisely from this combination, where domain expertise and artificial intelligence reinforce one another to unlock new operational capabilities. The result is Grapho, Intellico’s AI-powered design similarity solution, developed and refined in close collaboration with Phoenix to address the specific challenges of aluminum extrusion die design and knowledge reuse.
Grapho approaches the problem the way an experienced designer intuitively would — by recognizing shape and structure, not metadata. The system first learns a general sense of geometric similarity directly from Phoenix’s own archive of die profiles, using an unsupervised model to build an initial representation of each design. A contrastive learning stage then sharpens that representation, training the model to distinguish designs that are genuinely alike from those that only look similar at a glance — accounting for the rotations, mirrored orientations, and inconsistent filing conventions that accumulate across decades of archiving. A Graph Neural Network refines the process further, learning from the most informative comparisons in the dataset rather than from random samples, which makes the model measurably better at telling subtly different profiles apart.
Alongside the similarity engine, Grapho builds a transparent map of design families, so Phoenix’s engineers see not just a ranked list of matches, but a clear rationale for why two profiles are considered related — the transparency needed to trust the system’s suggestions rather than treat it as a black box.
That trust matters because the system’s effectiveness rests on two things working together: the engineering know-how Phoenix’s designers have built over more than fifty years, encoded in one of the sector’s richest technical archives, and the pattern-recognition capabilities of Intellico’s AI. Grapho does not substitute that expertise — it makes it instantly and systematically searchable, at scale.
In practice, a designer facing a new specification can query Grapho and scan the company’s entire design archive for the closest existing matches in few seconds, instead of relying on memory or manually paging through folders built up over years.
The numbers
In independent testing, Grapho matched or exceeded expert-level reuse identification — correctly surfacing reusable designs in more than 95% of simple profiles and over 80% of complex ones, comfortably ahead of manual and conventional search methods. The results encouraged Phoenix to move well beyond a pilot: having proven its value at the group’s core operations, the company is now extending the adoption of Grapho Design to its international subsidiaries, turning a single successful implementation into a shared capability across the wider group. Besides, the successful implementation of Grapho provided a positive internal case study to further promote the adoption of Industry 5.0 technologies inside the company.
For Technical Sales and Engineering teams, the same design-reuse capability translates into concrete, day-to-day gains that reach well beyond the design department. Engineering teams can also drastically reduce time to market through faster quotations for new projects, including preliminary production routing and cost estimation, bill of Materials rationalization, minimizing variants, duplicates, and non-standard items, and faster spare parts identification, improving serviceability and aftermarket responsiveness.
“For Phoenix International, the collaboration with Intellico represents a concrete step toward adopting advanced AI-based solutions. This allows us to optimize processes, leverage data, and support faster and more effective decisions.”
— Paolo Groff
CEO and Chairman of the Board, Phoenix International S.p.A.
The bigger picture
The challenges facing Phoenix are not unique to aluminum die extrusion. Across Engineer-to-Order and Make-to-Order manufacturing — from mechanical components to industrial tooling — companies sit on vast archives of past design work that traditional systems cannot search effectively. As product variety increases and customers demand faster turnaround, the ability to systematically reuse institutional knowledge, rather than starting from a blank page, will only become a bigger competitive differentiator. Explainable AI offers manufacturers a way to make that knowledge searchable, interpretable, and immediately actionable — turning decades of engineering history into a live asset rather than an archive.
