How to choose a robot for a business task

Choosing a robot for a business should begin with a task, not a product video. A machine can look capable in a controlled demonstration and still be wrong for your floor, load, people, network or operating model. The useful question is not “Which robot is best?” It is “Which complete deployment can perform this defined job under our real conditions?”

This guide gives you a decision tree before you compare the current Robotysys catalogue. It is a procurement framework, not a performance review. Manufacturer specifications tell you what a model is rated to do; only a representative trial can show whether the complete system works in your environment.

Step 1: write an observable task statement

Describe one job in a sentence that a neutral observer could verify. “Improve reception” is too broad. “Guide booked visitors from the reception point to four meeting rooms during staffed hours” can be tested. “Automate cleaning” is vague. “Scrub the mapped hard-floor route after closing, while a facilities employee handles exceptions” is usable.

Record five parts:

1. The object, person or information being moved or handled.
2. The start and finish of the task.
3. The required frequency or operating window.
4. The acceptable hand-off to a person.
5. The conditions that count as a failed attempt.

Do not combine unrelated jobs simply to justify the purchase. Reception, delivery, inspection and entertainment may require different hardware, interfaces and supervision. A narrow first deployment is easier to measure and easier to stop safely.

Step 2: classify the environment before the robot

The environment often removes more candidates than the task does. Walk the actual route and record:

• minimum clear width, door widths, lift dimensions and turning spaces;
• floor material, thresholds, ramps, stairs, drains, kerbs and drop edges;
• public traffic, children, animals, forklifts or other moving equipment;
• lighting, reflective glass, direct sun, dust, moisture and temperature;
• available power, charging position, Wi-Fi and mobile coverage;
• noise limits and any areas where cameras or microphones are inappropriate;
• fire routes, emergency access and places where queues form.

A product’s maximum payload, speed or ingress rating is not permission to operate anywhere below that limit. Those figures normally come from defined manufacturer conditions. Read the manual for the exact variant and ask which ratings apply simultaneously.

For example, Unitree describes the G1 as an early-stage humanoid platform and provides separate G1 manuals and tutorials. The existence of dynamic demonstrations does not make an unsupervised public route suitable. Similarly, a collaborative arm still needs an application-level assessment covering the tool, workpiece, fixtures, speeds and access—not just the arm’s label.

Step 3: choose the robot class

Use the smallest class that can do the job reliably.

This table is a shortlist, not a recommendation. A wheeled robot may be ideal on a stable mapped route and useless when a fire door remains closed. A robot arm may repeat a motion precisely but still require custom gripping, guarding, programming and commissioning before it produces useful work.

Step 4: separate the robot from the system

Ask suppliers to price the complete system in separate lines:

• exact robot variant and controller;
• batteries, chargers or docking station;
• end effector, trays, payload mount or sensors;
• software licences, cloud services and connectivity;
• mapping, integration and programming;
• delivery, installation and acceptance testing;
• operator and supervisor training;
• warranty, planned maintenance, spares and response arrangements;
• decommissioning or return obligations.

This is why a public robot price is an evidence anchor rather than the whole budget. The UK robot buyer’s checklist explains how to compare a quote line by line. If the requirement is uncertain, the rent, buy or RaaS framework helps decide who should carry early utilisation and maintenance risk.

Step 5: define human ownership

Every deployment needs named people, even when the machine is highly autonomous. Assign a business owner who decides whether the task still matters, an operational owner who controls day-to-day use, and a technical or supplier contact who can handle faults and changes.

Define who may start, stop, move, charge, clean, reset or reconfigure the robot. Specify who responds when it cannot complete a route, loses connectivity, detects an obstacle, drops a load or is approached by a member of the public. The Health and Safety Executive’s work-equipment training guidance makes the important point that training must fit the activity, equipment, existing competence and degree of supervision.

“The supplier will support us” is not an escalation plan. Record contact hours, response channel, what remote access is permitted and what the site should do while support is unavailable.

Step 6: choose the commercial route

Short managed hire suits fixed dates, public-facing work and cases where an operator is part of the service. A paid pilot suits repeatable tasks with uncertain fit. Purchase suits stable, recurring demand when your organisation can own integration, support and change control. RaaS may transfer parts of the service burden, but the contract needs clear uptime, maintenance, data, minimum-term and exit terms.

Do not turn an unproven pilot into a long contract because the demonstration went well. The demonstration proves that a demonstration can work. A useful pilot uses representative routes, loads, people and exception cases, with agreed stop criteria.

A one-page decision record

Before requesting quotes, complete this short record:

• Task: one observable sentence.
• Environment: route, floor, access, people, power and connectivity.
• Required result: measure and operating window.
• Human roles: operator, supervisor, support and decision owner.
• Known exceptions: what the robot will hand back to a person.
• Commercial route: managed hire, pilot, RaaS or purchase.
• Evidence required: demonstration, manual, site trial, references or acceptance test.
• Stop rule: the result or risk that ends the pilot.

That document is more valuable than a long wishlist of robot features. It gives manufacturers and integrators something they can answer honestly, and it gives your team a basis for rejecting an impressive but unsuitable machine.

When you are ready, share the task and site constraints for research guidance. Robotysys can point to relevant categories and evidence gaps; any shortlist, site scope or commercial quote must later be confirmed through an authorised supplier or integrator.