By Caitlin Taylor, Road Safety Manager, RoSPA
Ask a fleet manager whether they run self-driving vehicles and the answer will almost certainly be no. Ask whether their drivers use adaptive cruise control or lane-keeping assist and the answer is very likely yes. The gap between those two answers is where much of the risk sits today.
Automation is arriving at work in pieces, not all at once. It is in the lane-keeping on a company car, the driverless truck in a quarry, the delivery robot on a pavement and, since September, the supervised autonomous taxi a colleague can book in London (Wayve, 2026).
Each of these changes something about how people drive, how they work around vehicles and how risk has to be managed. This article draws on the automated vehicle fact files ScORSA developed with Anthrometric to look at where that change is showing up, what it does to people’s jobs and what employers can usefully do about it now (ScORSA and Anthrometric, 2025).
Where automation is showing up at work
The fact files sort workplace automation into six types. Each one turns up in a different part of working life.
-
In ordinary fleet vehicles.
This is where most organisations meet automation first. Adaptive cruise control, lane-keeping assist, blind-spot monitoring and emergency braking are now standard on many cars and vans. They help the driver. They do not replace the driver, who remains fully responsible throughout.
The trouble is that these systems can feel more capable than they are, and names like “Autopilot” do not help. The fact files offer a rule worth putting in every toolbox talk: if the driver must watch the road, it is not self-driving.
-
On private sites.
Ports, quarries, warehouses, depots and airport aprons are further ahead than public roads. Layouts are known, traffic is controlled and routes repeat, which suits automated haulage and yard vehicles. Companies such as Oxbotica and Volvo Autonomous Solutions already supply automated vehicles for mines, quarries and industrial sites. The fact files describe these sites as incubators for the technology, and for many employers this is where a driverless vehicle will first share space with their staff.
-
On campuses and business parks.
Low-speed electric shuttles run on fixed routes around hospitals, universities and business parks. Aurrigo’s Auto-Shuttle, for example, has been piloted in Cambridge, Milton Keynes and at the NEC. Most still carry a safety driver or onboard steward.
-
On pavements.
Small delivery robots, such as Starship’s in Milton Keynes and Northampton, already carry food and parcels at walking pace. The fact files point out that without careful planning they can block walkways and confuse people on foot, which makes kerb space a real issue.
-
By remote control.
Some vehicles are driven or monitored by a person in a control room, such as the remote valet service trialled in Milton Keynes. This is not self-driving. It swaps one set of risks for another: connection reliability, response times and operator workload.
-
As a passenger.
Autonomous ride-hailing has started in London with a trained driver on board to supervise (Wayve, 2026). Business travel policies will need to catch up with staff being able to book one.
What it changes about the job
For drivers, the work shifts from operating to supervising.
That sounds easier and often is not. Watching a system that works well nearly all the time is tiring in its own way, and people are poor at staying alert when nothing is happening, an effect known as vigilance decrement. New roles are appearing alongside the driver: onboard steward, remote operator, site supervisor.
For fleet managers, it is no longer only about driver behaviour.
The fact files argue that road risk management has to move towards assessing what each system can do, planning around the conditions it is built for and getting the interface between person and machine right. Putting a vehicle on a route outside those conditions is a planning failure, not a driver failure. A software update can also change how a vehicle behaves from one week to the next.
For people who work near vehicles, the hazards look different.
Automated vehicles are often quiet, may stop without warning and cannot make eye contact with someone waiting to cross. ScORSA’s guidance flags unexpected stops and silent operation as hazards to cover in site briefings (ScORSA and Anthrometric, 2025). Loading staff, banksmen and visitors need to know how these vehicles behave and who to call when one stops.
For incident investigation, the evidence changes.
After a collision, the system’s own data matters as much as the driver’s account. The fact files give an example of poor practice many organisations would recognise: no way to retrieve the system’s logs, so the report rests on the driver’s account alone. Employers need to know in advance what is recorded, who holds it and how to get it.
Automation moves risk around
The safety case for automation is real. Government cites human error as a contributing factor in 88% of collisions on UK roads (Department for Transport, 2026). A system does not get tired, glance at a phone or drive home after a 12-hour shift.
But removing one source of error does not remove risk. It changes where the risk sits, and some of the new places are harder to see.
- Handover. Asking a person to take back control at short notice is one of the best-documented problems in human factors. Attention fades on long, monotonous journeys, and a driver who has mentally checked out responds slowly or badly.
- Over-reliance. When people misunderstand what a system can do, they stop monitoring, start multitasking, or take risks they would not otherwise take because they expect the technology to compensate.
- Operating limits. Every automated system is designed for specific conditions: certain roads, speeds, weather or areas. Outside those limits it may hand back control or stop. Routes and schedules need to respect that.
- Cyber security. Connected vehicles can be exposed to hacking, GPS spoofing and sensor interference. That makes IT security a road safety issue.
- Mixed traffic. Automated vehicles will share roads and sites with human drivers, pedestrians, cyclists, motorcyclists and horse riders for years to come.
Boredom and overload are both hazards here. The fact files call these cognitive underload and overload: a person in the vehicle may have little to do for an hour and then face a sudden demand to take over. Shift patterns, cab design and training all need to account for that.
The legal position in brief
Most employers need three points for now.
First, nothing on sale in the UK today lets a driver hand over responsibility. The Automated Vehicles Act 2024 will allow authorised systems to become the legal driver while engaged, but most of it is not expected to be in force until 2027.
Second, the language is being tightened. From 7 January 2027, terms such as “self-driving” and “driverless” can only be used to market vehicles authorised to drive themselves (Automated Vehicles (Marketing Restrictions) Regulations 2026; Simmons & Simmons, 2026).
Third, the employer’s duty does not move. Under the Health and Safety at Work etc. Act 1974, risk must still be managed so far as is reasonably practicable, whoever or whatever is driving.
What to do now
Treat automation like any other new work equipment: assess the risk, train people, supervise, monitor and review. The steps below draw on the readiness checklists in the fact files. What they look like depends on where you are.
If your fleet has driver assistance, which is almost everyone:
- Find out what you have. List the assistance features across your vehicles, including grey fleet and hire cars.
- Fix the language. Use “driver assistance” and “self-driving” accurately in policies, inductions and toolbox talks.
- Train on limits, not just features. Drivers need to know when a system will not work, how it hands back control and when to override it.
- Write it into policy. Say when systems may be used and what is prohibited, such as using a phone while one is active.
- Use your data. Telematics and near-miss reports can show where attention is lapsing or systems are being misused.
If automated vehicles operate on your site, or soon will:
- Update the workplace transport risk assessment. Cover how the vehicles behave around people, other machinery and loading areas.
- Separate people and vehicles. Set routes, mark and sign them, and restrict where automated vehicles can go.
- Train everyone nearby, not only operators. That includes right of way, loading and what to do when a vehicle stops.
- Plan for failure. Decide what happens in bad weather, a lost connection or a breakdown, and who is allowed to move the vehicle.
- Agree the paperwork first. Settle with suppliers and insurers who is responsible for what, and who can access system data after an incident.
- Ask suppliers hard questions. What conditions is it designed for? What does it do at its limits? How has it been tested around pedestrians?
To support this, ScORSA has published Automated Vehicles: What it means for Occupational Road Risk Management, a set of fact files produced with Anthrometric. They cover six types of automated technology, each with a readiness checklist, alongside sections on the law, the effect on road risk and what automation means for groups from insurers to horse riders. Each section ends with questions to work through with your own team, such as which tasks, routes or sites would actually suit automation, and how to stop staff relying on it too much. They are written for employers who are not deploying anything yet as much as for those who are.
Keep hold of the responsibility
The technology will keep moving and so will the regulation. What stays fixed is that someone in every organisation is responsible for the safety of people who travel for work and of everyone they meet on the way. Automation may one day change who is driving. It will not change who has to manage the risk.
About the Author: Caitlin Taylor
Caitlin Taylor is Road Safety Manager at RoSPA and leads ScORSA, the Scottish Occupational Road Safety Alliance, a Transport Scotland-funded programme supporting more than 1,200 member organisations to manage occupational road risk.
References
- Automated Vehicles Act 2024, c. 10. Available at: https://www.legislation.gov.uk/ukpga/2024/10/contents
- Department for Transport (2026) Passengers one step closer to booking taxi and bus-style self-driving vehicles. 22 May. Available at: https://www.gov.uk/government/news/passengers-one-step-closer-to-booking-taxi-and-bus-style-self-driving-vehicles
- Health and Safety at Work etc. Act 1974, c. 37. Available at: https://www.hse.gov.uk/legislation/hswa.htm
- ScORSA and Anthrometric (2025) Automated Vehicles: What it means for Occupational Road Risk Management. Version 2.0.
- Simmons & Simmons (2026) Key updates to the UK’s self-driving regime. 28 July. Available at: https://www.simmons-simmons.com/en/publications/cms4ex26w00niuom8wicknvfp/key-updates-to-the-uk-self-driving-regime
- The Automated Vehicles (Marketing Restrictions) Regulations 2026 (SI 2026/733). Available at: https://www.legislation.gov.uk/uksi/2026/733/contents/made
- Wayve (2026) Wayve and Uber launch first-ever autonomous rides in the UK. 3 September. Available at: https://wayve.ai/press/wayve-uber-launch-autonomous-rides/
