Start with the process, not the robot's shape
A humanoid, mobile platform, manipulator or integrated machine is an implementation choice. First describe the workflow: input, required outcome, exceptions, safety constraints and current cost. A strong candidate is repetitive, measurable, constrained enough to control safely, and affected by risk or scarce labour.
Do not automate disorder. If routes change without rules, task data is unreliable and accountability is unclear, adding a robot only makes the underlying problem more expensive.
Split the portfolio into three horizons
Horizon one covers mature applications such as intralogistics, machine vision, inspection and repeatable manipulation. These projects should produce a near-term operational result. Horizon two contains adaptive systems that change behaviour from data or share space with people; they need deeper validation and observability. Horizon three includes exploratory general-purpose platforms and humanoids. Treat them as options to learn, not as guaranteed savings.
Give each horizon its own budget, stop criteria and acceptable risk. This prevents proven projects from being hidden behind costly experimentation.
Calculate the economics of the complete system
Include the robot, tooling, sensors, network, site mapping, WMS, ERP or service-desk integration, cybersecurity, training, spares and support. Compare more than human and machine hourly rates. Track process availability, manual intervention, quality, safety exposure, changeover time and the cost of downtime.
A solution scales when the next site requires configuration rather than a new engineering project. Evaluate site standards, interfaces and data quality separately from the demonstration.
Establish a robotics control plane
Every deployment needs a process owner, a technical owner, update rules, an action log and a safe degraded mode. Physical incidents cannot be reconstructed from an AI model log alone: controller events, sensors, fleet orchestration and safety systems must share a reliable timeline.
The portfolio is ready to grow when management can see business outcomes, engineering can see fleet health and security can restrict authority without stopping the entire operation.