For years, industrial automation adopted a comparatively secure system: high-volume manufacturing, lengthy product lifecycles, and robotic methods designed to repeat the identical course of hundreds of thousands of instances with minimal variation.
That mannequin helped remodel industries similar to automotive manufacturing, the place mounted manufacturing strains and extremely specialised robots delivered extraordinary effectivity at scale.
However manufacturing is altering. Product cycles are shorter, customization is growing, and factories are beneath strain to adapt extra rapidly to shifting demand and provide chain volatility. In lots of services, the normal economics of automation now not work as neatly as they as soon as did.
That altering atmosphere helps drive renewed curiosity in AI-powered robotics and software-defined automation – an space the place Alphabet-owned robotics firm Intrinsic is positioning itself aggressively.
Intrinsic, which emerged from Alphabet’s “moonshot manufacturing facility” ecosystem earlier than changing into a standalone firm beneath Google guardian firm Alphabet, is trying to construct a software program platform that makes industrial robots extra adaptable to altering manufacturing situations.
Slightly than changing present robotic {hardware}, the corporate focuses on including AI, machine imaginative and prescient, and clever automation layers designed to assist producers automate processes which have traditionally been thought-about too variable or uneconomical for standard robotics.
The technique displays a broader shift inside the robotics trade itself. After years of high-profile however commercially troublesome robotics experiments – together with Google’s former possession of Boston Dynamics and its long-running autonomous car efforts by Waymo – Alphabet now seems to be putting a extra deliberate emphasis on industrial and operational functions the place AI can clear up rapid manufacturing issues.
On this Q&A with Robotics & Automation Information, Stefan Nusser, chief product and industrial officer at Intrinsic, discusses why conventional automation is changing into “too costly and too rigid” in high-mix manufacturing environments, the place AI is already creating measurable worth on manufacturing facility flooring, and what Intrinsic has realized by its partnership with Foxconn about deploying AI-driven automation at industrial scale.
Nusser, whose background consists of management roles at IBM, Google, Willow Storage, and Fetch Robotics, argues that the way forward for manufacturing might more and more resemble software-defined infrastructure slightly than mounted manufacturing strains – with robotics methods changing into extra modular, reconfigurable, and adaptive over time.
Interview with Stefan Nusser
Robotics & Automation Information: You argue that conventional robotics is changing into “too inflexible” for contemporary manufacturing – what particularly has modified on manufacturing facility flooring that makes that rigidity a legal responsibility right now?
Stefan Nusser: Most automation right now is constructed for extremely standardized, low-mix processes, the place you do the identical factor many times. Automotive manufacturing is an effective instance in that the mannequin works when volumes are excessive and alter is proscribed. Nonetheless, a rising share of producing now not operates that means.
What’s modified is the extent of variability. You see smaller batch sizes, extra customization, and processes that evolve ceaselessly. In these environments, conventional automation turns into too costly and too rigid, as a result of it takes weeks to arrange for one thing that will solely run for a brief interval.
That’s why a big portion of labor stays unautomated right now, not as a result of it isn’t technically possible, however as a result of it hasn’t made financial sense. The chance now could be to automate processes that change constantly, which requires a special sort of flexibility.
R&AN: Many robotic distributors already declare flexibility by software program, imaginative and prescient methods, or reprogramming instruments – what’s essentially totally different about Intrinsic’s method?
SN: A whole lot of present options add flexibility onto methods initially designed for mounted, predictable processes. That works to a degree, however they battle when variability turns into the norm – totally different components, processes, and duties change ceaselessly. These methods nonetheless require vital handbook setup and re-engineering.
Intrinsic begins from the belief that change is fixed – enabling intelligently adaptive automation in high-mix, low-volume environments. The objective isn’t simply to make a single robotic cell simpler to reprogram, however to make automation simpler to create and handle day after day for non-experts on the store flooring.
In sensible phrases, which means utilizing AI and imaginative and prescient to deal with duties the place you possibly can’t realistically decide each element prematurely – like in electronics meeting, the place components differ and positioning isn’t excellent, and duties like cable dealing with or connector insertion are onerous to automate reliably.
At Intrinsic, we’re making these capabilities “prepared to make use of” and straightforward to mix in a single software program platform in order that they’re reusable throughout functions, slightly than rebuilt from scratch every time.
R&AN: From what you’re seeing in actual deployments, the place does AI really enhance efficiency right now – and the place is it nonetheless falling brief in manufacturing environments?
SN: AI is already creating actual worth in environments the place variability prevents automation and human staff are subsequently the bottleneck. Conventional automation works effectively for repetitive and predictable duties, however breaks down when issues change ceaselessly.
In high-mix, low-volume settings, AI permits us to deal with that variability. With basis fashions for machine imaginative and prescient, for example, we are able to interpret and manipulate objects with out figuring out their actual form prematurely – unlocking automation processes that have been beforehand uneconomical.
The place AI is rapidly catching up, is reliability. It isn’t 100% but however the newest fashions are extremely exact, correct and dependable – and enhancing at a gentle tempo. In manufacturing, the previous couple of proportion factors matter enormously, as a result of edge circumstances are the place downtime, high quality points, and complexity present up.
So the problem is ensuring AI is built-in simply as deeply as every other tooling or service – not simply mannequin efficiency in isolation. It’s about designing a full manufacturing system that may take care of exceptions safely and productively, usually with human oversight.
R&AN: Producers are beneath strain to justify automation investments – how does AI change the ROI equation in comparison with standard industrial robotics?
A: Historically, automation has made sense in very secure, high-volume environments the place you possibly can unfold the upfront value over lengthy manufacturing runs. The limitation has not been technical feasibility, however financial feasibility. If the method adjustments too usually, the setup and reprogramming value can outweigh the profit.
AI adjustments that equation by lowering the trouble required to deal with variability. As an alternative of rebuilding or closely reprogramming and retooling the system each time a course of adjustments, you can also make automation extra adaptable.
That opens up a special class of alternative. You’re now not creating single-purpose automation that’s to be amortized over the lifespan of a single product; you’re creating reusable “multi-purpose” automation that may be re-configured for a special product when wanted.
This enables the funding in automation expertise to be amortized over the lifespan of the automation {hardware}, which will be so long as 7-10 years.
R&AN: The partnership with Foxconn suggests large-scale validation – what have you ever realized to this point about deploying AI-driven automation in high-volume, real-world factories?
SN: One of many greatest classes is that even in high-volume factories, actual world manufacturing isn’t as uniform as folks think about. There may be nonetheless numerous variability between components, processes, product generations, and line situations.
That’s precisely when clever robotic methods, powered by AI, brings modularity and flexibility to advanced manufacturing processes – that beforehand weren’t reasonably priced or manageable by producers.
The power to reconfigure and subsequently repurpose manufacturing capability on-the-fly opens up new alternatives to share infrastructure throughout a number of merchandise, react rapidly to sudden shifts in demand and subsequently cut back the necessity for larger stock ranges of completed product.
R&AN: There’s usually a spot between what producers want and what robotics firms are promoting – the place do you assume the trade remains to be getting it incorrect?
SN: I believe the most important mistake the trade makes is beginning with the expertise as an alternative of the issue. There’s an inclination to generalize too early in constructing one thing that may do many issues, after which assume the worth will comply with.
You see that particularly within the push towards very general-purpose methods, the place the expectation is that one resolution can handle a variety of use circumstances. That creates numerous pleasure, but it surely additionally makes it more durable to outline a transparent start line and ship actual buyer worth, in the best way they want it delivered.
There’s totally different worth inherent to totally different approaches. One other path is to give attention to a small variety of issues, and go deep with the intention to show that real-world worth will be created. Solely after that do they begin to generalize and discuss platforms.
R&AN: Trying forward, do you see AI making present robots extra helpful, or in the end changing present industrial methods altogether?
SN: Within the close to time period, it’s about making present methods extra helpful. The {hardware} is already there, the true query is methods to make it versatile sufficient to deal with variability. That’s the place AI is having an instantaneous impression – extending what present methods can do.
At the moment’s industrial robots are cheaper and higher than ever. With AI they may turn into much more versatile and usable, together with for staff with out robotics expertise or experience. Robotic methods are additive, giving staff extra help in additional adaptive methods, and making present and new methods extra versatile and versatile.
Over time I believe it results in a special mannequin altogether. As an alternative of mounted manufacturing strains designed for a single product, you progress towards extra versatile, software-defined environments. Similar to how manufacturing is changing into extra like a knowledge middle – the place you might have a pool of sources, and what you produce is configured by software program and might scale up or down dynamically, as wanted.
However that shift will occur step-by-step. It begins with fixing particular functions the place the worth is evident, then constructing from there towards broader orchestration throughout workcells, strains, and finally bigger manufacturing facility operations.
