Electronics Era: Why will factories and warehouses be the first major testbeds for embodied AI?
Ankur Yadav: Factories and warehouses are likely to become the first major testbeds for embodied AI because they offer structured environments where physical tasks can be clearly defined, measured and improved. Picking, sorting, palletising, material movement and repetitive handling are already established automation use cases, making them natural starting points for more adaptive systems.
Humanoids can be viewed as the next extension of industrial automation, particularly where operations are designed around human workers and existing infrastructure. The test, however, will not be whether a robot can perform a task once, but whether it can do so reliably, safely and economically at scale. With more than 300 customers across 25 countries, our experience at Addverb has shown that Physical AI will scale fastest where it addresses measurable operational challenges rather than remaining a technology demonstration.
Electronics Era: How is automation shifting from programmed systems to adaptive, intelligent robots?
Ankur Yadav: Automation is moving from robots that simply follow instructions to systems that can interpret their surroundings and decide how to act. Traditional industrial robots are highly effective when conditions are predictable, but they typically depend on predefined movements, rules and layouts. Physical AI adds perception, reasoning and adaptive control, allowing robots to respond when objects, people or operating conditions change. The next generation will increasingly learn from real-world actions and use that experience to predict sequences of movements rather than execute isolated commands.
Over the last couple of years, we have been exploring this evolution at Addverb as well, building on our experience in warehouse automation and mobile robotics and extending our work into quadrupeds and humanoids. This is driven by a broader shift in how we think about robotics, from machines that execute predefined tasks to systems that can perceive, reason and adapt to what is happening around them. The goal is not simply to automate another task, but to build machines capable of making better physical decisions in real operating environments.
Electronics Era: How are real-world data and industrial environments accelerating Physical AI?
Ankur Yadav: Physical AI improves through contact with the physical world because industrial environments contain variability that simulations cannot fully capture. Objects differ in shape and weight, people move unpredictably, layouts change and edge cases emerge during live operations. Every deployment can therefore generate valuable information on movement, objects, environments, failures and recovery. That data can be used to improve perception, planning and control, creating a continuous learning loop between deployment and model development.
For more than three years, we have been exploring Physical AI at Addverb and building the data pipelines needed to learn from real-world robotic operations. Today, we process around 1,000 hours of operational data each week, and we expect that volume to scale significantly as our deployments grow. This creates a continuous learning loop between deployment and model development, helping improve perception, planning and control.
Electronics Era: What could human–robot collaboration look like over the next decade?
Ankur Yadav: Human-robot collaboration over the next decade is likely to be defined by task allocation rather than replacement. Robots will increasingly handle repetitive, physically demanding or hazardous activities, while people focus on judgement, supervision, exception handling and decisions that require context.
Humanoids could accelerate this model because their form factor is designed to operate in spaces already built around human movement, tools and workflows. This does not mean every task should be given to a humanoid. The right system will depend on the economics, variability and safety requirements of each process.
The opportunity, as we see it, is to create human-robot workflows where machines extend human capability, improve safety and raise productivity rather than simply substitute for labour.
Electronics Era: What does India need to build to compete in the global Physical AI race?
Ankur Yadav: India’s ambition in Physical AI should extend beyond assembling robots to building the underlying technology stack. That means capabilities across AI models, perception, sensors, controls, actuators, embedded systems, electronics and advanced manufacturing. Localisation of critical components will matter because import dependence can constrain cost, supply resilience and the ability to customise systems for global markets. India also needs stronger links between startups, manufacturers, universities and testing environments where robots can learn from real industrial conditions. Addverb’s localisation journey and global expansion illustrate the value of developing products in India while designing for international requirements.
Electronics Era: How is Addverb moving embodied AI from concept to commercial reality?
Ankur Yadav: We see embodied AI as an extension of the industrial robotics journey we have been on, rather than as a standalone technology experiment. Our experience across warehouse automation, mobile robotics and intelligent systems has given us access to real operating environments where reliability, safety and integration are critical.
Our move into humanoid robotics is the next phase of that journey. In 2026, Addverb introduced Elixis-W, a wheeled humanoid designed for industrial applications, while further humanoid development is underway. The focus is on moving from demonstrations to deployable systems that can perform useful tasks consistently and integrate into existing workflows.
The goal is not simply to build a robot that can demonstrate what is technically possible, but to develop systems that can create measurable value in real industrial environments.
