Are we overengineering AI brains for logistics? (Filics × Servo7 × Sandwater)

​Physical AI is the last layer of a multi-layer stack. The customer isn't buying the layer — they're buying the full stack, and they want it to solve their problem predictably, consistently and as simply as possible.

​This roundtable brings together two founders deploying robots in warehouses and containers today, who claim the AI needed to deliver what the customer actually wants is the easiest part of the stack.

​It took Filics five years to make a robot drive straight — reliably, safely, in a real facility, with real pallets and real forklifts around it. Servo7 unloads loose-loaded shipping containers at 40°C in the summer, and their customers don't ask what model runs on the arm. They ask whether the container gets emptied. Every time.

​The thesis we'll put on the table: the value in robotics lives in the unglamorous layers — hardware, sensors, safety systems, motion control, determinism — not in the vision-language-action model bolted on top. The customer wants the simplest solution that works repetitively, not the smartest one. And the player who sells that solution to the end customer, and guarantees its performance at 3am, captures the lion's share of the wallet.

​Join us if you're building hardware and agree, if you're building models and disagree, or if you're an investor still trying to figure out who ends up on top.

​Speakers:

​Gregor Kolls — CEO, Filics.

​Pieter Becking — Co-Founder & CEO, Servo7.

​Chloe Fu — Sandwater invests €2-10m checks into Seed-Series B start-ups across the Nordics, Germany and Benelux, with a focus on hardware and hardware/software combinations across Energy, Industry and Health.