Buying Driverless Buses in Britain
Exploring Britain's first autonomous bus procurement specification
The first specification for buying autonomous bus services is here, and it describes a driverless service that may need more staff per passenger than human-driven buses.
That specification is the Automated Public Transport Services draft, published for consultation by Transport for West Midlands (TfWM) and Solihull Council. It is a 53-page template for how local authorities should procure autonomous bus services. It covers everything from wheelchair ramps to black ice to preventing anti-social behaviour.
What makes this document unusual is both its timing and ambition. Local authorities almost never publish procurement specifications before national regulatory guidance exists. The rules for approving driverless passenger services have not been finalised. TfWM and Solihull published anyway.
The authors want local authorities to have a voice in shaping this technology. This specification is one contribution; their recent paper on the local authority perspective to connected and automated mobility is another. And they are aiming high. If implemented as written, this specification would set standards for autonomous buses above those we currently expect from human-driven services.
The implications extend beyond transport. This specification touches procurement practice, labour economics, surveillance technology, accessibility standards, and how local government navigates emerging tech. Anyone interested in how Britain adopts new technology has reason to pay attention.
How the specification works
This specification is a template for how UK councils should buy autonomous bus services, the first of its kind, doubling as a wish list for what local authorities want from this technology.
Despite the broad title ‘Automated Public Transport Services’, the specification focuses on buses and shuttles, with a focus on roads and no mention of rail or trams. It’s designed to help local authorities and transport authorities purchase automated services, keeping all details in one place.
The template comes pre-populated with multiple choice options (for example, selecting maximum service speeds of 15, 30 or 60 mph), but is structured for easy modification. This opens the door for case-by-case negotiations between operators and transport authorities.
What follows is a close reading of what the specification assumes, what it demands from operators, and what it reveals about its authors’ strategic positioning.
The Specification’s Assumptions About Technology and Delivery
The many functions of a bus driver
The document reveals that matching the service a human driver provides requires significant technology beyond self-driving software. A bus driver does far more than drive. They deploy wheelchair ramps, manage capacity and judge when to stop boarding passengers, reroute in response to traffic events or road closures. They also play a security role, intervening against anti-social behaviour or refusing entry to passengers carrying dangerous items. All of these were explicitly mentioned as service requirements within the specification.
The specification also misses a few considerations that are usually delivered by the driver. Drivers detect mechanical issues based on steering feel or engine sounds and send vehicles for checkups accordingly. They also handle unexpected situations like giving directions to lost passengers, alerting emergency services when someone falls ill, or communicating with traffic police during incidents. The specification leaves open which of these functions must be automated and which can be dropped from the service entirely.
Frontier technologies beyond self-driving
The specification expects continuous CCTV footage both inside and outside vehicles as part of this service. It calls for footage to be kept for 28 days and made accessible to the transport authority. It is implied that these CCTV capabilities would monitor or intervene with unexpected or dangerous passenger behaviour and aid law enforcement investigations should a crime be committed on board.
This adds massive technical complexity. Any provider would have to deliver both an autonomous solution and an integrated system of onboard cameras with real-time image recognition. How precise does image recognition need to be to confirm specification requests as specific as whether wheelchairs are appropriately anchored before departure?
The consortium model
The specification calls for a single bidder referred to as the “operator,” but the operator may be “a system manufacturer, a transport operator, a hybrid, collaboration, or joint venture, or an as yet undefined body.” The authors explicitly state that “who does what” within the fulfilment of the requirements is for the market to decide, recognising that future commercial models may differ.
This gives the market flexibility on what consortiums and joint ventures to build when bidding. Joint operations in autonomy are standard. Waymo combines its software with vehicle manufacturers, ride-hailing platforms and maintenance partners across multiple cities.
If this specification or a similar consortium model becomes the norm, it will provide a commercial pathway for new firms to become regular partners for autonomous public transport. Will this support local and innovative businesses, or will larger players like Stagecoach, Siemens Mobility and Cisco take continuous slices of these services?
The Specification’s Demands on Operators
Accessibility beyond legal minimums
TfWM and Solihull could have stated that accessibility requirements are set by existing legislation and left it there. Instead they go meaningfully beyond that. Their expectations include vehicles yielding for guide-dog-led pedestrians, induction loops at every priority seat, and exploring “options (if any) for those without smartphone access or indeed the unbanked.” Their ambitions are packaged in one quote: “Services must ensure, as far as is reasonably practicable, that no demographic is dissuaded from [access].”
Can you procure feelings of safety?
An important UK safety principle appears within this specification. It is called “as far as reasonably practicable” (AFARP), a cousin of “as low as reasonably practicable” (ALARP). ALARP means you must reduce risk as much as you reasonably can, unless further reduction would cost wildly more than the safety benefit. It is widely used in the transport sector, but implementation often depends on context, regulatory culture and sector precedent.
The specification does something unusual: it applies the UK “reasonable practicability” idea to how safe the service feels, not just how safe it is. It does this while leaving the practical definition open. The “definition and process [is] to be agreed with the TA.” The council is not just buying a safe service. It is buying a service that people feel safe using, and it reserves the right to decide what “feels safe” means and how operators prove it.
The first wave of pilots and winning tenders will likely set the tone. Once one authority-operator pair settles on a workable evidence pack (survey measures, complaints categories, incident handling standards, transparency reporting), it becomes the obvious reference point for the next procurement. Not because it is law, but because procurement loves a template.
This also gives the transport authority a lever to push back on glossy bid claims. If a bidder sells “people will feel safe” as a headline benefit, the authority can respond: show me what you will measure, how often, what triggers action, and what you will do when the numbers go the wrong way.
The risk is the flip side. If the transport authority does not define acceptable evidence early, bids either become optimism contests, get bogged down in debates, or everyone lives with uncertainty. At that point you are no longer arguing about vehicles or safety standards. You are arguing about whether “reasonable” safety feelings have been delivered.
Three possible futures emerge. Does “perception of safety” become a minimum standard that operators must hit? A scoring feature that earns nice-to-have points? Or a living KPI that changes as trials mature? These single-line details can delay or hasten months of procurement.
The oversight problem
The whole point of autonomous buses is removing the driver. But the specification layers in extensive human oversight. It calls for remote supervision with “rapid response” capability, physical human response within minutes of an alert, a Service Panel with emergency services and public representatives, potential co-location of operator staff at transport authority premises, transport authority approval before software updates, and a human-driven backup contingency service that must be maintained and tested regularly.
Many of these requirements make sense in isolation. Remote supervision catches edge cases. Physical response reassures passengers. The Service Panel ensures accountability. Co-location can build trust. Software approval prevents untested changes. Manual backup covers failure modes.
Taken together, they paint a picture of a “driverless” service that may require more staff per thousand passengers than existing services. Staff monitoring screens. Staff on standby to respond. Staff attending panels. Operator staff potentially sitting in council offices.
The specification does not address this tension directly. It sets the requirements and leaves the economics to bidders. At what point does the cost of oversight erode the savings from removing the driver?
The Specification’s Strategic Position
Written before the rulebook exists
The authors acknowledge that the specification is published “ahead of the publishing of the UK Automated Passenger Service (APS) Permitting guidance.” The APS guidance will be the plain-English rulebook for how a driverless passenger service gets approved to run.
Given that this specification for local procurement is written ahead of national regulation, it “may therefore be subject to significant revision.”
Why invest so much effort into a document that will probably need rewriting? Potentially because its existence can shape what “good” looks like before national guidance lands. The authors are explicitly asking for feedback. This specification can give TfWM and Solihull more weight in the debate. Many of these considerations could end up in the final APS guidance or other elements of the Automated Vehicles Act regulatory environment.
A consultation, not a conclusion
This is the sixth version, published explicitly as a draft for wider consultation. The public feedback deadline was 12 December 2025, with a final specification targeted for Spring 2026. It is a draft inviting contributions from academics, industry and the public. But it is also on track to become the reference document for local authorities nationwide.
The authors are honest about what they do not know. They make no fixed assumptions about the shape of the mobility industry. They are open to modifications. Operators will negotiate, and pragmatic compromises will arise. But TfWM and Solihull have done thinking others have not, and that thinking is now on the table for anyone who wants to shape it.
Conclusion
This specification reveals something often left unsaid: autonomous public transport will not be cheaper if we demand the same service we have now.
When you replace one worker with a consortium of firms, continuous CCTV, real-time image recognition, rapid-response staff, service panels and backup human-driven vehicles, you may find you have traded a known cost for an unknown one. The specification does not address this directly. It sets high standards and leaves the economics to bidders.
But the implication is clear. If implemented as written, this specification would raise the minimum acceptable standard for public transport beyond what human-driven buses currently provide. That is not cost-cutting. That is quality improvement. The closest parallel may be body-worn cameras on police officers. That technology improved accountability and service quality without reducing headcount. Autonomous buses, under this specification, could follow the same pattern.
This may be the right approach. First-generation autonomous services might need to be better than what they replace, not just cheaper, to earn public trust. But other local authorities may have different priorities. Some councils will want autonomous buses precisely because they cost less. This specification does not give them a template for that. TfWM and Solihull have set a high bar. Whether other authorities adopt it, adapt it, or write their own remains to be seen.
The specification is clear and thorough. It cites every regulation a future service would need to comply with. It does the thinking other councils have not. The next local authority attempting this will not start from zero.
The conversation is early. National guidance has not landed. Operators will negotiate. Clauses will be revised. But the terms of the debate are being set now, by whoever shows up to shape them. TfWM and Solihull showed up. The next public round of consultation is likely to come soon. The door is open for anyone else who wants to contribute.
