ibbw-bertrandt provider logoOpen for submissions

Engineering Knowledge Copilot for Bertrandt

Bertrandt is seeking an AI-powered Engineering Knowledge Copilot that captures scattered and tacit engineering know-how and makes it instantly accessible in daily work. The goal is to preserve expert knowledge, accelerate onboarding of young engineers and support engineers during design and project tasks, validated in a Venture Client PoC.

Objective

Business Goal:

Preserve and share engineering expertise across the organization and accelerate the onboarding and productivity of engineers, so that expert know-how is not lost and both experienced and newly hired engineers can access it easily in daily work.

POC Goal:

Within the Venture Client framework, validate a working copilot in a controlled pilot that:

  • Demonstrates a genuinely low-effort way to capture knowledge — from existing documents and directly from employees in the flow of work, ideally by voice
  • Provides a usable, easy-to-operate copilot accessible via a web interface or Microsoft Teams
  • Delivers reliable, source-referenced answers on selected engineering use cases (e.g. design guidelines, checklists, release/approval processes)
  • Shows measurable user acceptance among the German and international pilot user groups
  • Uses (automotive) cockpit module development as the lead use case, with metal as the material focus; extension to further modules (e.g. door) and materials is a possible next step after the PoC
  • Delivers a working MVP that pilot users in Germany and at the international site in Morocco can test

The PoC should deliver a working pilot within approximately 3 to 6 months and is expected to focus on a limited set of use cases and a defined pilot user group before broader rollout and extension. Given the emphasis on reusable building blocks, a provider that can start from an existing solution can enter the PoC with a working tool rather than only a concept.

Long-Term Goal:

Establish a scalable engineering knowledge copilot that Bertrandt can continue to extend and, ideally, further develop itself without deep IT expertise. Potential extension stages include role-aware answers, self-updating from new project learnings captured via text or voice, running alongside a live CAD session, and connecting to external benchmark databases.

Solution Requirements

Functional

  • Answer engineering questions based on internal knowledge, e.g. on design guidelines, and support daily engineering tasks — for example general design guidelines for metal and plastic parts (draw depths, radii, draft angles, tool/slider directions), with (automotive) cockpit module development as the lead use case
  • Support the creation and revision of DFMEA and documentation, summarize specifications and help create status reports
  • Make capturing knowledge as effortless as possible: ingest existing documents (PowerPoint, Word, PDF, Excel) and, crucially, let employees contribute tacit know-how with minimal effort in the flow of work — by simply speaking (voice input) as well as by text
  • Provide reliable, source-referenced answers with traceability so outputs can be verified and hallucinations avoided
  • Search and reference general legal and homologation requirements (e.g. ECE regulations, NCAP criteria) from generally accessible sources alongside internal knowledge
  • Never invent compliance-relevant content: for laws, norms and guidelines the system must state when no source is found instead of generating an unsupported answer
  • Support German and English, including for international teams
  • Be accessible in daily work via a web interface or Microsoft Teams

Technical & Security

  • Deployable within Bertrandt's cloud environment, currently Microsoft Azure, with hosting in the EU
  • Focus on internal data, while allowing enrichment with external data where useful; no uncontrolled exposure of internal knowledge
  • Built on existing products or reusable modules where possible rather than a full greenfield development
  • Designed to avoid vendor lock-in, so Bertrandt can continue to operate and extend the solution without depending on a single provider or deep IT expertise — ideally extendable by Bertrandt itself with little or no programming knowledge
  • Able to pass Bertrandt's procurement, compliance and IT-security requirements (specific GDPR / IT-security criteria to be confirmed by procurement)
  • Show compatibility with Bertrandt's internal AI framework (“AI stack”), which is oriented toward maximum openness and compatibility; details will be shared under NDA at PoC start
  • Initial assessment (“first guess”) of the requirements for achieving TISAX certification and the potential challenges, particularly with regard to future scalability.

Operational readiness

  • Deployable as a working pilot within approximately 3 to 6 months
  • Able to involve Bertrandt's internal AI/IT team early (e.g. at PoC kick-off) and to show the team how to further train and extend the copilot

Nice-to-haves

  • Role-aware behavior: recognize (or let the user set) the user's role and adapt terminology and required answer precision accordingly
  • Voice interaction in addition to text
  • Self-updating from new knowledge captured during projects
  • Ability to run alongside a live CAD session and provide design feedback
  • Ability to connect to external benchmark databases as a later extension
  • Ideally also with image recognition, to run analyses (Even better would be a direct interface to the CAD system, to "see" directly whether something is incorrect)
  • Extension to customer-specific design guidelines (e.g. OEM-specific CAD modelling requirements) after the PoC
  • Analyze CAD-derived images or sections (e.g. screenshots provided as PDF/PowerPoint) and check them against applicable design guidelines

Out of scope

  • A solution requiring lengthy interviews that pull engineers out of productive work to capture knowledge
  • A pure off-the-shelf product with no adaptation to Bertrandt's knowledge and use cases
  • A closed solution that Bertrandt cannot continue to operate or extend if the provider is no longer involved

Example of such an application

  • Employee: I'm currently designing a fender and changing the radius of the headlight wrap from 5 mm to 3 mm. Is this still manufacturable at all?
  • AI: Could you provide me with the following information: sheet thickness, material, and approximate draw depth?
  • Employee: (provides the information)
  • AI: Yes, it should still be manufacturable. This has also been implemented in the following vehicle projects: POxxx, …
    (ideally with a reference to the corresponding component or even to the CAD file)

Partner Requirements

The partner should:

  • Brings an existing productized solution or reusable core components (not a pure greenfield custom-development project), and can also advise on how to best capture the knowledge and turn it into agents
  • Headquartered in Europe, with EU-based hosting and no dependency on US-only infrastructure
  • Able to work in German and English (international teams work in English)