Even at 99% reliability, an industrial robot can still require so many human interventions that the economics do not work.
The real test goes beyond impressive demonstrations: can the technology operate reliably in production, be deployed quickly and scale economically?
At the EUVC Summit & Awards Show 2026, Tomaz Stolfa, Co-Founder and CEO of Sunrise Robotics, joined Patrick Murphy, Partner at Tapestry VC, which is an investor in the company, to discuss what it takes to bring robotics from promising technology to working production lines.
Europe's manufacturing challenge demands a different approach
Tomaz pointed to a looming workforce challenge: close to a third of Europe’s manufacturing workforce is expected to retire over the coming decade, according to reporting by Fortune.
Traditional automation is well suited to high-volume manufacturing, where predictable production justifies expensive, customised systems.
But Tomaz estimates that two-thirds of European manufacturers operate in high-mix, low-volume environments, where changing production requirements make conventional automation harder to justify.
For these businesses, flexibility matters as much as productivity.
As he explained, manufacturers want to move people into higher-value work while handing repetitive tasks to machines. Whether a system is marketed as AI-driven matters far less than whether it delivers that outcome.
Deployment speed is part of the product
An average industrial automation project takes 12 to 18 months, according to Tomaz. Sunrise is reducing that process to weeks.
Rather than treating each installation as a separate engineering project, the company has developed a process that captures data from human operators before deploying robotic cells prepared for similar workstations.
Once positioned and connected, a cell can begin operating within one shift at a sufficiently similar workstation.
Crucially, Sunrise has separated data acquisition, robotic cell preparation and hardware production. It designs the machines but does not manufacture them itself, allowing different parts of the deployment process to scale in parallel.
For Tomaz, solving robot programming alone is insufficient if the rest of the installation process remains slow and complex.
“If you just figure out a better way to programme robots, that doesn't make a dent because you still have to do a bunch of other things.”
Reliability and real-world deployment matter more than demonstrations
Tomaz estimates that current leading vision-language-action models achieve around 80% to 85% robustness. Even 99% would leave too many interruptions in an industrial environment. He argues that factory applications need reliability of 99.9% and beyond.
Industrial robots must also operate at productive speeds while meeting demanding physical and safety requirements. This makes the choice of robotic hardware and physical design as important as the intelligence controlling it.
Patrick highlighted another critical distinction for deep-tech companies: actually delivering working products.
Sunrise reached a factory production line approximately 18 months after its first hire. At the time of the conversation, Tomaz said the company still had robotic cells operating on production lines daily.
“You're actually deployed in factories and deployed in days.”
The implications extend to both company building and venture investment. Hardware businesses need operating models that support physical deployment, alongside investors prepared for long development horizons.
Tomaz emphasised that Sunrise has deliberately built its investor base around long-term alignment, describing company building as a journey that could take 15 years or more.
Three questions for assessing industrial robotics companies
The conversation offers three practical questions for founders, operators and investors thinking about what makes an industrial robotics business scalable:
Do the economics hold up in production? Assess reliability, human intervention requirements and whether automation genuinely reduces repetitive work.
Can deployment scale without becoming a bespoke project? Examine how data collection, preparation and hardware production are organised, and whether installation timelines can be consistently shortened.
Is the technology operating in real factories? Look beyond demonstrations to actual production deployments, daily operation and evidence that the company can replicate its approach.
These distinctions help separate technical potential from the ability to build a scalable industrial business.
Listen to the full EUVC Summit conversation for more on Sunrise's approach to factory automation, robotics deployment and building a hardware company for the long term.
