Spend any time with CTOs and the chat is almost always about AI. Copilots, chatbots, analytics pipelines, document understanding – the digital side of the business is getting plenty of attention. But in most of those conversations the AI stops at the edge of the screen while the warehouse, factory floor and inspection bay still run much as they were a decade ago.
Physical automation has always existed. But it has sat behind a wall of specialist expertise, capital and long lead times that most IT and software teams could not see past. That wall is now coming down. The question for technology leaders is whether they treat that change as part of their AI strategy, or leave it as somebody else’s problem.
From niche tech to mainstream platform
My background is in software rather than robotics. After two decades building software teams and technology businesses, I started to look more closely at the intersection of AI and physical automation. What I found was familiar: a specialist domain starting to look like a platform that generalist development teams can use.
Cloud computing is the obvious precedent. Before cloud, running infrastructure meant capital expenditure, specialist skills and lengthy procurement cycles. The arrival of platforms such as AWS, Azure and Google Cloud abstracted that complexity into services that a small team could consume in hours rather than months. The systems did not become simpler; they became accessible through a different model.
Robotics is moving in the same direction. ROS2, the standard framework for programming industrial robots, is powerful and open source but difficult to learn. Making progress has meant understanding a large stack of concepts and tools, which has kept robotics in the hands of specialist roboticists who are in short supply. For many organisations, that has been enough to keep robots on the ‘future plans’ slide.
But recently, platforms have emerged on top of ROS2 that provide browser‑based environments, integrated simulation and SDKs in common languages. Instead of assembling their own mix of simulators, dashboards and custom scripts, development teams can work in one place and treat robots as just another class of endpoint in the architecture. These capabilities are already in production; what is lagging is recognising them as part of mainstream AI and automation strategy.
The robot team you already employ
As perception and planning increasingly rely on AI models, the bottleneck in robotics is shifting away from low‑level control towards data management, orchestration and integration – areas where existing software and IT teams are already experienced.
A developer can now log into a web‑based environment, spin up a simulated industrial robot and experiment with behaviours before any hardware is ordered. Large language models can generate ROS2 code from natural‑language descriptions. Developers review that code, run it in simulation and refine it as they would any other component. The code sits in the same repositories, goes through the same review processes and is deployed through the same pipelines as other software.
This changes who can own automation strategy. Rather than building a separate robotics function, organisations can draw on engineers who already understand their systems and data, and treat robots as part of the same stack rather than as isolated projects.
Rethinking how you buy robots
Traditionally, each robotics deployment has been treated as a bespoke project. You hire specialist roboticists, engage systems integrators, commit capital to hardware and integration, then discover over time whether the automation meets expectations. It is a familiar process for anyone who remembers pre‑cloud infrastructure projects.
A platform‑based approach alters both the economics and the sequence of decisions. Development starts in simulation. Use cases are validated before hardware is purchased. Organisations can run small pilots, discard the ones that do not deliver and scale up the ones that do. Robots become standardised endpoints running defined behaviours, rather than one‑off builds.
Consumption models are evolving too. Some providers now offer robot‑as‑a‑service, where customers pay for hours of operation instead of owning the asset outright. Combined with sim‑first development, this changes the risk profile of automation projects in a way that mirrors the move from owned hardware to cloud computing.
For IT leaders who lived through cloud procurement debates – build versus buy, capex versus opex, open versus proprietary – these patterns should be recognisable. Organisations that engaged with cloud early, in a measured way, gained advantages that compounded over time. The same potential exists in physical automation.
The ‘last mile’ of your AI projects
The highest‑profile robotics deployments tend to be in warehousing and logistics, where autonomous mobile robots move goods between zones and robotic arms pick from shelves. But for many organisations, the more immediate opportunities sit at the boundary between digital programmes and the physical environment.
For predictive maintenance, for example, many manufacturers have invested in sensors, data pipelines and AI models to forecast equipment failures. In practice, these initiatives often culminate in dashboards and alerts on control‑room screens. When an alert sounds, a technician still walks the floor, inspects the asset and performs a standard intervention. Detection has been automated; the response remains manual.
Robots can handle parts of that response. A mobile platform equipped with a camera and basic tooling can be dispatched to a machine, collect visual or sensor data and carry out a simple inspection or reset. The workflow from model output to physical action can be orchestrated by the same software teams that built the predictive maintenance system.
A similar story appears in warehouse operations. Many facilities now use AI‑driven demand forecasting and sophisticated warehouse management systems to decide what to pick, when and to where. Execution can still often be manual however, with staff walking long distances to carry out those decisions. Autonomous mobile robots and robotic picking systems can take the instructions those systems already generate and turn them into physical movement on the warehouse floor.
Organisations have digitised much of the data layer through sensors, telemetry and analytics; but the connection to consistent physical action is still developing. In sectors facing acute labour shortages such as warehouses, ports, construction, brownfield manufacturing and field service, combining human expertise with robotic assistance is becoming a practical response to current constraints.
A strategic choice, not a technical one
Every significant technology shift creates a period in which early adopters can build advantages that are difficult to replicate later. Cloud, mobile and data platforms all followed that pattern. Organisations that engaged early, took time to understand new models and built internal capability now operate differently as a result.
Physical automation is entering that phase. The factors that kept robotics at arm’s length from mainstream IT – the need for scarce specialist skills, the capital intensity of hardware purchases, the complexity of bespoke integration – are being reduced by open frameworks, platform layers, simulation environments and new consumption models.
For CTOs, the question is not whether physical systems will become part of IT strategy, but when and on what terms. If competitors are quicker to connect their AI investments to physical systems, closing the loop from predictive maintenance alerts to automated inspection, or from demand forecasts to in‑facility logistics, they will accumulate operational advantages over time.
The practical move is to treat physical automation the way you would treat any emerging technology. Understand the procurement models. Run a contained pilot that links one existing digital programme to a simple physical workflow. Build internal capability in the teams who already know your systems. If you are still treating robotics as a one‑off patch at the edge of operations, that is now a choice, not a technical inevitability.
Nick Thompson is co‑founder and CEO of Sheffield-based OLO Robotics.
- AI in Supply Chain
- Data & AI