[Adobe Stock]
The gap between market enthusiasm and operational reality is what Thomas Brown sees from his vantage point as senior director, engineering and technology, digital commerce at Jabil. Brown has spent more than 20 years across Jabil, Flex, and a range of software platforms, giving him a long view of where hardware ambition collides with production reality. Robotics follows an adoption arc similar to many other emerging technologies. Getting pretty good output is easy in pilots. Getting excellent quality work at scale is a different ballgame. “Making something work 80% of the time is straightforward,” Brown says. “Getting to 99.9% reliability is exponentially more difficult. And what happens when a robot breaks down at 2:00 a.m.?”
In the conversation below, Brown walks through what actually stalls robotics deployments at scale, where humanoids fit in the near-term calculus, and why manufacturing expertise belongs in the design conversation from the start.
What’s the bottleneck between the current wave of robotics interest and actual adoption?
Thomas Brown
Brown: Selling a robot is not just selling a robot. It’s about providing an entire system solution: the function the robot performs, integration into the customer’s workflow and processes, IT systems, physical space, and all the systems needed to service and maintain it.
There’s also a large gap in reliability between a demo and real, continuous operation. The environment is another challenge. Our world is designed for humans, which presents all kinds of edge cases for robots. The areas where deployments have succeeded are places where environments can be standardized: warehousing, logistics and manufacturing.
Beyond the technical, there are non-technical reasons for reluctance. Unions, safety culture, public sentiment and real or perceived job displacement all weigh on decision-makers. And even as robots become more affordable, the investments needed are still high. Given the uncertainty around performance and reliability, companies are hesitant.
Ultimately, the constraint is economic. Automation solutions are purpose-built for a narrow window of application. Minor changes to a product, line configuration or volume can invalidate the automation and require costly redevelopment. That’s exactly why humanoids generate so much interest. If the same platform can be redeployed across changing products and workflows, the amortization math works over many years instead of a single product cycle. That changes the investment case entirely.
Which robotics use cases are closest to scaled deployment today, and which ones are still overbuilt for the near-term problem?
Brown: The successful integrations share repeatable motion, bounded variance and clear ROI math. That includes material movement via AGVs and AMRs, dense storage retrieval, palletizing and depalletizing. ASRS systems carry higher complexity and cost, but the value is unambiguous in operations with scale and variability.
Humanoids are the most discussed category and the furthest from broad commercial viability. What’s interesting is where the gating constraint actually sits. The hardware has advanced faster than most people realize. The harder problem is the policy layer, involving vision-language-action models, world models and the edge compute to run them with the latency and reliability a physical system requires. A step change there would shift the deployment timeline for humanoids more than any hardware advance.
Can you provide a high-level overview of how Jabil is using robotics, both as a provider of contract manufacturing services and engineering services?
Brown: Robotics shows up in two places at Jabil: in what we manufacture, for both robotics customers and non-robotic customers that require complex, custom automation systems; and in how we engineer those products and solutions. Jabil has a large fleet of retail inventory robots deployed in the world, and we continue to design and develop new autonomous robots to solve real-world problems for a wide range of customers.
On the engineering services side, one of the most impactful applications is software validation. Modern devices are software-defined, and software complexity introduces bugs that can ripple through the system and manifest in actuation and device behavior. Validation has to scale with that complexity: more test cases, executed with precision, quickly and in parallel.
We’ve deployed this in the FinTech space, where security and quality requirements are unusually high. Software engineers issue a pull request, the code is deployed to a fleet of robotic test cells that execute regression testing in parallel, and coverage scales without scaling human hours. The result is faster release cycles with higher confidence in quality and security outcomes.
Can you walk through an example where a robotics product looked viable in pilot but became difficult to scale, and a case where deployment proved transformative? What can we learn from the difference?
Brown: Robotic products carry massive complexity. Mechanical, electrical, electromechanical and software systems all have to coordinate to perform a useful operation, and the path from prototype to manufacturable product is consistently underestimated.
A common failure pattern happens when a product engineering team focuses first on “can it work?” then on “can it work reliably?” and only then engages a manufacturing partner. By that point the architecture is locked. Executives have made commitments to investors and customers based on a commercialization timeline that assumed the hard manufacturing problems were already solved. They are not. Design changes and supply changes may be needed to hit a viable price point and yield, and it is a difficult position to navigate.
The contrast is a recent engagement where the customer brought us in deliberately to value-engineer a platform already in the field: targeted design-for-manufacturability work, supplier rationalization and yield improvement on an existing bot. The result was meaningful cost-out and reliability gains without disrupting the deployed installed base.
The lesson is engagement timing. Robotic products are too tightly coupled between design and manufacturing for those conversations to happen sequentially. Manufacturing expertise belongs in the design conversation from the beginning, treated as an architectural input rather than a procurement step.
