If your social feeds are like ours, they’re probably filled with robots of every shape and size performing feats that would have been unimaginable a few years ago: knocking off chores with untiring efficiency, dancing with influencer IShowSpeed in China, running half-marathons, lifting heavy objects, sparring with black belts, and so forth.
These demos are incredible, but their value goes far beyond tricks and novelties. The real value will arrive when robots extend digital agency into the physical world. As sectors like healthcare and construction face persistent labor shortages, robots have the potential to augment human capabilities and support critical tasks in these and other fields. Consider the numerous roles where increased support could be invaluable: nurses, teachers, firefighters, disaster response personnel, and paramedics, to name a few. Why wouldn’t we want a world with fewer shortages in these vital roles, allowing human professionals to focus on the most complex and human-centric aspects of their work?
Think about robots that respond to signals from sensors and other sources in high-risk fire areas and coordinate with drones to prevent property damage, or construction robots that augment human crews to build cheaper, better housing in areas where persistent shortages impede broader access. Robots might play a growing role in elder care, ensure faster service in retail and hospitality during peak periods, and help maintain both public and private spaces with significantly greater scale than currently possible (“Hey assistant, could you please have them send one of the roadbots to fix that pothole on Dahlia and 14th Avenue?”).
While the potential for robots to augment human capabilities and support critical tasks is immense, realizing this future will require careful consideration of crucial factors, like physical safety measures, the need for predictable and reliable operation, the nuances of psychological human-robot interaction, and the broader ethical and data security implications of increasingly autonomous machines.
Salesforce Futures has been engaging with robotics experts inside and outside Salesforce to learn more about what we can expect in the years ahead. These conversations have revealed three technological and strategic forces that business leaders should consider now to position their organizations for the incoming robotics wave.
1. Convergence of digital and physical intelligence
Breakthrough improvements in AI are helping machines make sense of and interact with the world around them. In the past, robots struggled to make sense of their environment and perform complex tasks because they couldn’t perceive and interpret information effectively. Now, advancements in deep learning are accelerating robots’ ability to see, understand, and respond to their surroundings. As Salesforce researcher and robotics PhD Juan Carlos Niebles told us in a recent Salesforce Futures Roundtable, AI’s perceptive capabilities have rapidly evolved, bringing us closer to the long-sought goal of creating robots with more generalized physical intelligence. “AI perception advanced very quickly once we realized how fast deep learning would evolve,” he said. “Deep learning and end-to-end training are arriving for robots.”
In our conversations with Niebles, he points to new Vision Language Action models as akin to “LLMs for robots.” Nvidia’s Isaac GROOT N1 shows where that might go: An open, fully customizable foundation model that promises to give general-purpose humanoids a modular library of reasoning and skills. Amazon’s just-launched Vulcan robot already proves the utility of this concept on the warehouse floor. Vulcan’s force torque sensors allow it to “feel” objects and safely pick or store the inventory in a fulfillment center.
Deep learning and end-to-end training are arriving for robots.
Juan Carlos Niebles, Research Director, Salesforce
Combining generalized reasoning with real-world perception and touch results in a flywheel for learning that mirrors the compounding gains already seen in LLMs. This makes physical AI very relevant for both Salesforce and our customers. We’re especially interested in how robotics — in combination with data, intelligence, applications, and agents — will help companies connect with their customers and deliver next-gen service and experiences.
Think about robots that interact with customers in the physical world and share back what they learn, contributing valuable data that helps the broader robotic fleet get smarter. In the same roundtable Niebles attended, MuleSoft solutions engineer Viktoriya Kotik pointed out that once context comes from different data sets, the diversity of inputs will make it much easier to give robots what they need to learn. Extending on this concept, Nvidia’s Director of AI Jim Fan points out that simulated data in the form of “digital twins, cousins, and nomads” can help deliver these broader data sets faster and cheaper than we previously thought possible.
Advancements like these could pave the way for highly versatile, general-use humanoid robotics. Humanoids that not only capture our attention and imaginations because they look like us — they also promise cheap and useful labor augmentation in a world full of pressing shortages that show no signs of easing. These generalized capabilities may even eventually extend to domestic settings, as illustrated by Tesla’s Optimus robot concept, which can apparently now perform many household tasks.
2. Enterprise service transformation demand
While humanoid robots capture the imagination, many valuable near-term applications are not getting the attention they deserve. As we survey the opportunity landscape through the lens of Salesforce and the customers we serve, we get most excited about practical applications and use cases for enterprises, particularly in service. For example, Bert Legrand, part of the team spearheading robotics efforts inside Salesforce, has shown a demo where a Boston Dynamics Spot robot performs tasks like checking a pressure gauge and medical supplies, dispatched and managed via Slack, Agentforce, and Field Service. The example points to a world where asset management, automation, and enterprise intelligence converge to deliver better, more efficient services across industries.
We believe service use cases will multiply as robots gain increased capabilities. Salesforce EVP & GM of Field Service Taksina Eammano is particularly focused on the near-term, practical utility of robotics in field service use cases.
“The next frontier of field service isn’t about replacing human workers, but augmenting workforces with intelligent robotics that can handle repetitive, dangerous, or hard-to-reach tasks. We’re seeing breakthrough applications in infrastructure inspection, remote maintenance, and predictive service interventions that can dramatically improve safety, efficiency, and response times,” Eammano told us. “These use cases show how Salesforce can use our platform (including products like Field Service, Agentforce, Mulesoft, and Slack) to extend digital labor into the physical realm.”
The next frontier of field service isn’t about replacing human workers, but augmenting workforces with intelligent robotics that can handle repetitive, dangerous, or hard-to-reach tasks.
Taksina Eammano, EVP & GM of Field Service, Salesforce
CRM insights promise to be an important part of personalizing and elevating every interaction, including those involving robots. Consider a hotel maintenance scenario that transforms routine upkeep into relationship-building gold. Before the arrival of Carla Rodriguez, a repeat guest, sensors detect a humidity spike in room 412, but can’t explain the cause. A compact field service bot runs by for a quick scan. The robot’s thermal camera spots a damp footprint spreading behind the minibar — likely a condensate line leak — and, on the way out, its vision model flags a coffee stain sure to displease Carla, whose profile notes she really values not only air quality, but cleanliness. Both issues flow into the same service case, enabling a human technician to arrive with the right gasket and a paint touch-up ticket. Carla arrives, once again, to a room that meets her standards.
3. Physical world implementation complexity
While there are reasons for enthusiasm about LLM-like scaling, many of the experts we spoke to, including Legrand and Niebles, were quick to point out that numerous challenges remain before generalized humanoids go mainstream in every conceivable context.
First, the famous adage “hardware is hard” has endured because, despite AI and other innovations, delivering good hardware that customers love remains insanely difficult. Practical challenges like battery power, cost, maintenance, and fleet coordination present real obstacles for engineers trying to scale robotics. Even the seemingly simple way a human toddler learns to interact with the physical world is still incredibly difficult for robots to replicate. Additionally, making robots capable of operating in new environments without significant engineering presents a major hurdle.
Second, delivering these robots safely presents a whole other set of challenges. Most of the humanoids we see on our social media feeds look lithe and agile, but are surprisingly heavy in person (i.e., you really don’t want one falling on you). Recently, a video of a humanoid robot windmilling its arms wildly without awareness of its surroundings went viral as viewers imagined the terrifying prospect of being anywhere near those flailing metal appendages.
Finally, human-robot interaction (HRI) is complex and nuanced, and those who study this multi-disciplinary field point out that assumptions about how robots might interact with humans are often proved incorrect in practice. The subtle psychological and physiological responses humans have to robotic movements reveal the intricate nature of these interactions. For example, jerky and unpredictable robot motions can trigger a fight-or-flight reflex in humans. A 2022 study by Jeanne Kirsschner and others found that when a robot unexpectedly approached human volunteers, almost one in three humans flinched and behaved unpredictably, raising collision risk even at low speeds. Designing the relationship between humans and robots across sectors and contexts will require thoughtful attention, iteration, and refinement over time, even as learning and intelligence continues to advance.
How should business leaders prepare for the robotic future?
Our conversations haven’t changed our view that robots will soon become widespread. However, we now believe their widespread use might look less dramatic than what people currently imagine.
We envision specialized, purpose-built robotics designed for specific use cases — like manufacturing and service — proliferating over the next three to five years. These robots, some of which already exist, will likely scale at a more measured pace, giving engineers time to learn and refine, and allowing for the evolution of human norms.
We envision specialized, purpose-built robotics designed for specific use cases — like manufacturing and service — proliferating over the next three to five years.
But we should offer a caveat. As the rapid advancements over the last two years in AI have taught us, we can’t assume that tomorrow’s technologies will advance like yesterday’s. The powerful combination of advancements across fields and the potential of AI agents to accelerate the pace of change means that business leaders can’t stand pat and wait for the robotic future to unveil itself.
The most successful organizations will be those that use robotics not just to boost productivity, but to deliver new value and better customer experiences. At Salesforce, we’re hosting design sessions around the world to imagine the future, prototype robotics applications with real-world utility, and partner with our customers to get ahead of the coming wave. Now is the time for business leaders to start imagining what physical AI might mean for the business and its customers, and experimenting with those possibilities so that when the future comes, the company is ready.