Summary of "Ferdinand Schockenhoff - Capgemini Invent - Product development 2030+"

Product development 2030+

(Ferdinand Schockenhoff, Senior Manager, Digital Engineering & R&D, Capgemini Invent)

Speaker: Ferdinand Schockenhoff (introduced by Jan).
Focus: Automotive-centered examples presented as transferable to other industries.
Goal: Make product development faster, cheaper, higher-quality and multi-functional through AI and digitization.

Increasing time-to-market and cost pressure.
Growing software content and electrification in products.
New competitors and business models.
Current engineering work is often manual (Office tools, phone calls) and not machine-actionable — agents cannot operate unless development artifacts are digitized and standardized.

Systems engineering foundation
- Model effect chains: capture functional and physical interactions between components and roles (who talks to whom).
- Digitize these models so they are explicit rather than implicit in engineers’ heads.
Add simulations and structured data
- Describe component or black-box behavior via simulations, databases and workflows.
- Ensure each part has defined inputs/outputs and physical boundaries.
Deploy AI agents on top
- Only after steps 1–2 can autonomous agents (multi-agent systems) meaningfully design, integrate and make decisions.
- Example: a battery-geometry agent + vehicle-integration agent + cost/approval agent working together.

Do foundational modernization in parallel with concrete AI use cases.

Standards & methods
- Use consistent modeling languages/methods across effect chains (e.g., SysML-type approaches).
IT platforms
- Prefer standardized, scalable platforms that provide data continuity and an integrated UX over one-off tools.
Structured data
- Choose a function- or system-oriented end-to-end data architecture and enforce it.
Organizational alignment
- Rethink responsibilities and create roles that enable automated/agent-driven steps.

Process selection
- Identify high-impact, high-automation, repeatable engineering tasks (avoid “nice-to-have” chatbots).
Define clear end-to-end boundaries
- Specify explicit start and end points for the automated task.
Data check
- Confirm available input data and required outputs (e.g., CAD positions, engineering BOM → collision-free geometry + interference report).
Stepwise decomposition
- Break the process into subprocesses/tasks; implement each as an agent where sensible.
Enablement
- Train and onboard engineers and process owners; adjust organization and governance.
Scale
- Chain agents into multi-agent systems; reuse and recycle agents across use cases.

Inputs: data models (positions), engineering BOM.
Agent tasks:
- Run geometry/clearance checks.
- Generate interference report.
- Pass results to integration, cost, and approval agents.
Output: collision-free geometry validated for manufacturing.

Build foundational tooling: platforms, sandboxes, infrastructure, and libraries of reusable agents/components.
Embed governance, legal, compliance, and AI regulation into the strategy.
Invest in upskilling across product development so agents are actually used and produce business impact.

“Agentic AI is arriving quickly; organizations that delay risk falling behind.”

Combine foundational modernization with quick, high-impact use cases to demonstrate value and steer transformation.
Focus on the processes engineers actually perform (value‑stream oriented).
Without digitization and structured data, AI agents cannot be effective.
Act fast to capture advantage from emerging agentic AI capabilities.

Ferdinand Schockenhoff — Senior Manager, Digital Engineering & R&D, Capgemini Invent
Moderator / introducer: Jan (name not given in transcript)
Organization referenced: Capgemini Invent