Robotics in Hong Kong Care Services: Workforce Support, Independence and the Limits of Automation

A worker in a residential care home spends part of each shift moving supplies between floors. Another repeatedly assists with physically demanding transfers. A rehabilitation professional can supervise only a limited number of therapy sessions each day. An older resident wants more opportunities to practise movement without waiting for a staff member to become available. These are the kinds of practical problems in which robotics may have a meaningful role in Hong Kong care services.

They also sit within a much broader gerontechnology strategy. Hong Kong has spent several years expanding technology adoption across elderly and rehabilitation services, supported by the Innovation and Technology Fund for Application in Elderly and Rehabilitation Care and a growing ecosystem of technology developers, welfare organisations, residential services and community programmes. The wider Hong Kong Ageing, Long-Term Care & Community Support Knowledge Hub examines how these developments connect with ageing, disability services, workforce capacity, community support and quality governance.

Robotics is one part of that transformation, but it needs particularly careful analysis. A robot that transports equipment is very different from a device that interacts socially with a person. A rehabilitation robot is different again. The operational, ethical and workforce consequences depend upon the task being automated, the person using the technology and the extent to which human judgement remains involved.

The strategic question is therefore not how many robots a care organisation can introduce. It is which tasks should be automated, which should be augmented by technology and which should remain fundamentally human.

Hong Kong has created substantial infrastructure for gerontechnology adoption

The Social Welfare Department established the Innovation and Technology Fund for Application in Elderly and Rehabilitation Care in December 2018 with an initial HK$1 billion. Its purpose is to improve quality of life for older people and persons with disabilities while reducing pressure on care staff and carers through the procurement, rental and trial of suitable technology products.

The scale has grown significantly. By the end of 2025, around HK$910 million had been allocated and approximately 2,100 elderly and rehabilitation service units had been supported to procure or rent more than 27,000 technology products. An additional HK$1 billion was injected into the fund in 2024-25, and its scope was expanded to enable suitable products to be lent for use in people’s homes.

The fourteenth tranche opened in October 2026 for eligible private and self-financing residential care homes that do not receive Social Welfare Department subsidies. This continued funding demonstrates that technology adoption is no longer being treated as a short experimental programme.

Robotics sits within a wider product landscape that also includes monitoring systems, mobility aids, transfer equipment, rehabilitation technologies and smart-care infrastructure.

That context matters because robotic care should not be understood as a standalone innovation category. Its value depends on how it connects with people, buildings, digital systems and service workflows.

Robotics already covers very different functions

The term “care robot” can create misleading images of a humanoid machine attempting to replace a care worker. Hong Kong’s real gerontechnology landscape is considerably broader.

Technology already used, trialled or explored across elderly and rehabilitation settings includes:

• robots transporting equipment and materials;
• intelligent robots conducting scheduled patrols;
• robots supporting virtual communication and visits;
• companion and interaction robots;
• robotic rehabilitation and exercise technologies;
• automated mobility and transfer-support systems; and
• emerging humanoid and embodied-AI applications.

The differences are fundamental.

A transport robot primarily redesigns logistics. A rehabilitation robot supports repeated therapeutic activity. A companion robot enters the person’s emotional and social environment. A patrol robot may gather information about people and spaces.

Each requires a different assessment of safety, privacy, dignity, competence and benefit.

Kwu Tung North shows how robotics can be designed into care environments

The Multi-welfare Services Complex in the Kwu Tung North New Development Area provides an important Hong Kong example of technology being integrated into the design of care infrastructure rather than added later.

The complex includes residential care homes for elderly people and persons with disabilities. Gerontechnology provided within the environment has included contactless vital-sign monitoring, bed-exit and fall-detection systems and intelligent robots capable of scheduled room patrols and supporting virtual visits.

The strategic value of this model lies partly in integration.

Technology can be difficult to deploy in existing buildings if corridors are narrow, connectivity is inconsistent, charging areas are unavailable or workflows were designed entirely around human movement.

New facilities can consider robotics alongside architecture, electrical infrastructure, digital connectivity and staffing from the outset.

This connects robotics with the wider design of automated workflows. The objective should not be to insert a machine into an unchanged service. It should be to decide whether technology allows the service itself to operate differently.

Scenario: a transport robot changes the work around care

A large residential care home receives regular deliveries of linen, continence products, meals and other supplies across several floors. Care workers frequently interrupt direct support to collect equipment or move materials between storage areas and resident areas.

The organisation introduces a transport robot capable of completing selected internal deliveries.

The initial business case focuses on labour saving. After implementation, managers discover that direct staffing numbers do not fall substantially. Instead, the main benefit is different: workers are interrupted less often while supporting residents.

A care worker helping an older person prepare for lunch can finish the interaction rather than leaving to collect equipment. Night staff spend less time moving routine supplies. Supervisors can plan replenishment more consistently.

However, new tasks emerge. Routes must remain clear. Staff need to know how to stop the robot safely. Somebody must monitor charging, faults and maintenance. Deliveries still need human oversight where items are urgent, sensitive or clinically important.

The organisation therefore changes how success is measured. It stops asking how many staff hours the robot has “replaced” and instead measures interruptions to care, delivery reliability, staff physical workload and time available for direct support.

This is a stronger model of robotics because it evaluates technology through service outcomes rather than machine utilisation alone.

Workforce benefit is likely to come from task redesign rather than worker replacement

Hong Kong faces persistent workforce pressure across elderly and rehabilitation care. Robotics is therefore understandably discussed as part of the response.

But describing robots as replacements for care workers oversimplifies both the technology and care itself.

Many care tasks contain physical, administrative and relational components simultaneously. A worker assisting somebody to transfer from bed is not simply moving weight. They may be observing pain, discussing how the person feels, encouraging participation and adjusting support in response to subtle changes.

Automation may reduce part of that physical burden without replacing the interaction.

The stronger workforce opportunity lies in redesigning tasks so technology handles activities that are repetitive, physically demanding or logistically inefficient while staff retain functions requiring judgement, communication and relationship.

This makes robotics closely connected with safe staffing and deployment. A robot may change what staff do, but it does not remove the organisation’s responsibility to ensure sufficient competent human support remains available.

Reducing physical burden can have consequences for retention

Care work can involve repeated pushing, lifting, transfers, bending and transportation of equipment.

The effect of robotics on workforce sustainability may therefore be more important than simple productivity.

If suitable technology reduces musculoskeletal strain, physical exhaustion and repetitive manual work, it may help experienced staff remain in the workforce longer.

This is particularly significant in an ageing workforce.

Organisations should therefore evaluate robotics against outcomes such as injuries, fatigue, sickness absence and worker experience rather than only throughput.

The Predictive Workforce Risk Module can help organisations examine whether workforce instability, sickness, workload and continuity pressures are concentrated in particular services before deciding whether technology could address part of the underlying problem.

Rehabilitation robotics offers a different model of augmentation

Rehabilitation technology demonstrates especially clearly why robotics does not need to mean replacement.

Robotic systems can support repeated movement, structured exercise and measurement of performance. They may enable a person to practise more frequently while rehabilitation professionals use information generated by the system to adapt intervention.

The professional role remains essential.

A machine can repeat a movement accurately. It cannot independently decide what outcome matters most to the individual, whether fatigue has become clinically significant or whether repeated practice remains meaningful to the person.

Robotics can increase the amount or consistency of practice. Human rehabilitation professionals interpret what that practice means.

Scenario: rehabilitation moves beyond the therapy session

A person recovering from a neurological condition attends a Hong Kong rehabilitation service. He needs repeated upper-limb exercise, but therapist capacity limits the number of fully supervised sessions available.

The service introduces a robotic rehabilitation system that guides structured movements and records performance.

Initially, the technology is used only during therapist-led sessions. As confidence develops, selected exercises can be completed with lighter supervision while staff remain available.

The system records range of movement and completion patterns, giving the therapist additional information when reviewing progress.

However, the data does not become the person’s rehabilitation goal. He wants to regain sufficient control to prepare simple meals independently, not merely improve a movement score.

The therapist therefore links robotic exercise to practical activity and adjusts the programme around the person’s own objectives.

Progress is assessed through both machine-generated measures and everyday function.

The technology increases practice capacity, while professional judgement keeps rehabilitation connected to meaningful independence.

This reflects the wider principle behind equipment and assistive technology for people with physical disabilities: the value lies in what technology enables the person to do, not simply in the sophistication of the device.

Robots supporting independence need to be judged from the person’s perspective

Robotics may also support daily living outside formal rehabilitation.

Devices could assist mobility, retrieve objects, provide reminders, support exercises or connect people with others remotely.

For somebody living with physical disability, the right technology could reduce dependence on another person for specific tasks.

That can have profound implications for autonomy.

Waiting for another person to move an object, operate equipment or provide routine assistance may seem minor from an organisational perspective. For the person, repeated dependence can shape their entire day.

This is why robotics should be evaluated through person-centred technology.

The strongest test is not whether the device performs the task. It is whether the technology increases the person’s control, convenience, privacy or participation without introducing disproportionate new risks.

Home deployment changes the governance equation

The expansion of Hong Kong’s Innovation and Technology Fund to products suitable for household use is strategically important.

Technology that works inside a staffed residential service may behave very differently in somebody’s home.

In a care home, staff can notice faults, resolve connectivity problems and remove equipment from use. At home, responsibility may sit with an older person, a disabled person or a family carer.

This means deployment needs to consider:

• whether the person understands how the technology operates;
• who responds if the device fails;
• whether the home environment is physically suitable;
• whether connectivity is reliable;
• how charging and maintenance will be managed;
• whether family involvement is expected; and
• what happens when the product is no longer appropriate.

The move from institution to home therefore cannot consist simply of lending the same equipment to a different address.

Support arrangements must travel with the technology.

Companion robots raise more complex questions than logistics robots

Hong Kong’s gerontechnology landscape has included companion robots for several years.

Such technologies may support stimulation, activity, conversation or emotional engagement. They can be particularly attractive in services supporting people who experience loneliness or cognitive impairment.

Yet the ethical questions are more complicated than those surrounding a machine that transports supplies.

A companion robot may appear responsive, emotionally expressive or socially present. People may attribute understanding or intention to the device that it does not possess.

This does not automatically make the technology inappropriate.

Objects, music, animals, games and digital media can all provide meaningful engagement without possessing human consciousness.

The important distinction is whether the technology enriches the person’s life or substitutes for relationships they would otherwise have received.

A robot used during an activity programme alongside attentive staff is different from a robot introduced because the service no longer has time for conversation.

Scenario: a companion robot produces unexpected benefit

An older woman with dementia living in a residential care home becomes increasingly withdrawn during large group activities. Staff introduce a companion robot during several smaller sessions because she appears interested in tactile and responsive objects.

She begins interacting with the device and talking more frequently while a care worker sits with her.

The team initially considers the robot itself to be the intervention. Observation suggests something more complex. The device appears to provide a focus that makes conversation with staff easier.

The care plan is therefore adapted. The robot is used selectively as one engagement tool, alongside music, familiar objects, reminiscence and individual conversation.

Staff record whether it improves mood and participation and stop using it when the woman appears uninterested.

Importantly, the device is not used automatically whenever staffing is pressured.

The organisation treats the outcome as increased engagement rather than reduced demand for human interaction.

This is particularly relevant within dementia support, distress and meaningful activity. Technology can contribute to engagement, but its meaning depends on the individual response rather than the novelty of the device.

Acceptance should not be assumed because technology is available

Robotics often attracts strong reactions.

Some people find robots interesting, useful or entertaining. Others may find them confusing, intrusive or frightening.

Age alone does not predict acceptance.

An older person with significant technology experience may be enthusiastic about robotics, while a younger disabled person may reject a device that feels stigmatising or unnecessarily visible.

Choice therefore matters.

Care organisations should avoid presenting robotic systems as mandatory simply because funding has been secured or equipment has already been purchased.

Where a robot directly interacts with the person, their preferences should shape deployment wherever practicable.

Procurement success should not be confused with implementation success

Hong Kong has created substantial financial mechanisms to support technology acquisition.

That solves one important barrier: capital cost.

It does not automatically solve implementation.

A robot can be purchased, delivered and technically operational while contributing little to care.

This happens when technology is selected without sufficient understanding of the workflow it is intended to improve.

Organisations therefore need to define the operational problem before selecting the product.

The sequence should be need, intended outcome, workflow design, product selection and evaluation—not product first and purpose afterwards.

The Digital Transformation Readiness Assessment can help organisations examine whether infrastructure, workforce capability and governance are sufficiently mature before introducing complex technologies.

Maintenance determines whether robotics remains useful after the pilot

Robotics introduces a practical issue often underestimated during innovation projects: machines require maintenance.

Batteries deteriorate. Sensors become misaligned. Software updates change functionality. Replacement components may become unavailable. Suppliers may leave the market.

This creates a long-term cost that is different from initial procurement.

An organisation considering robotics needs to understand who provides technical support, expected equipment life, response times for faults and what happens when the product reaches end of support.

Where technology is integral to care, downtime also requires contingency planning.

If a robotic transfer device becomes unavailable, staff need an alternative safe method. If a delivery robot stops working, essential supplies still need to reach residents.

Technology should therefore reduce operational vulnerability rather than create a new single point of failure.

Robotic safety extends beyond conventional equipment checks

Robots that move through shared environments introduce distinctive safety considerations.

A mobile robot may encounter residents using walking aids, people with visual impairment, wheelchairs, dropped objects or rapidly changing environments.

A rehabilitation device may exert physical force. A transfer-support technology may interact directly with a person’s body.

Risk assessment therefore needs to reflect actual use rather than supplier specifications alone.

Near misses are especially important.

If staff repeatedly need to intervene because a robot struggles with a particular corridor, threshold or crowding pattern, the absence of injury does not mean the system is operating safely.

Those interventions are safety data.

Workforce competence cannot stop at initial training

New technology is often accompanied by supplier training at launch.

Several months later, trained staff may have left, new workers may rely on informal demonstrations and uncommon functions may be forgotten.

Robotics therefore requires continuing digital workforce capability.

Workers need to understand not only how to operate a device but when not to use it.

Competence includes recognising malfunction, responding safely to unexpected behaviour and understanding the limits of automation.

Managers also need sufficient knowledge to challenge claims that a robot is improving productivity merely because utilisation figures are high.

Scenario: technology adoption creates hidden work

A rehabilitation service introduces several robotic devices intended to increase productivity.

Usage statistics appear encouraging. The machines are booked for most of the working day.

However, staff report that each session now requires additional time configuring equipment, cleaning sensors, resolving minor errors and entering duplicated information into another system.

The apparent productivity gain exists only if machine operating time is measured without the surrounding work.

Management maps the full workflow and discovers that some tasks can be simplified. Equipment settings are standardised where clinically appropriate, digital records are better integrated and responsibility for routine technical checks is clarified.

After redesign, staff time falls while patient access to rehabilitation remains high.

The lesson is that robotic productivity should be measured across the whole service process.

A device that performs one task faster can still make the service less efficient if the surrounding workflow becomes more complicated.

Robots generate data as well as activity

Many modern robotic systems record information about usage, performance, location or interaction.

Rehabilitation devices may record movement and repetition. Mobile robots may map physical environments. Interactive systems may process voices, images or behavioural information.

This connects robotics directly with information governance.

Organisations need to know what data is generated, whether it identifies individuals, where it is stored and whether suppliers can access it.

The fact that equipment appears mechanical rather than digital does not remove privacy obligations.

Increasingly, the robot is simply the physical interface to a much larger data system.

Embodied AI will make the boundary more important

The next generation of robotics is likely to combine mechanical capability with increasingly sophisticated artificial intelligence.

Hong Kong’s Digital Policy Office has already launched a “Robots in the Community” pilot programme designed to give citizens direct exposure to humanoid “Smart Bot” technology and embodied AI through community events.

The initial programme is primarily about public awareness and digital inclusion rather than formal care delivery. That distinction is important.

Humanoid robots attracting public interest should not be presented as established care-worker replacements.

However, embodied AI indicates the direction in which technology is moving. Robots may increasingly be able to navigate environments, process speech, respond dynamically and perform a wider range of physical tasks.

Care governance therefore needs to prepare before capability becomes commonplace.

Automation becomes more sensitive as the robot moves closer to the person

A useful governance principle is that scrutiny should generally increase as robotics moves from environmental tasks towards direct human interaction.

Automating transport of linen creates relatively limited ethical complexity.

Supporting physical rehabilitation requires stronger clinical oversight.

Assisting transfers requires robust safety controls.

Monitoring a person introduces privacy questions.

Providing companionship enters the relational environment.

Making decisions on behalf of the person would create an entirely different level of concern.

This suggests that robotics should be governed according to consequence rather than novelty.

A visually impressive humanoid robot may actually create less risk if it simply welcomes visitors than an inconspicuous automated system influencing important care decisions.

Data should show whether robots improve care rather than merely whether they are used

Once organisations have invested in technology, there can be pressure to demonstrate success.

Usage counts are easy to produce.

They are also weak evidence.

A robot completing 2,000 journeys says little about whether staff workload improved. A rehabilitation device completing thousands of exercise repetitions does not prove that people became more independent.

Useful evaluation should connect technology with outcomes such as:

• changes in staff physical workload or injuries;
• time released for direct care;
• rehabilitation and functional outcomes;
• service-user experience and choice;
• reliability and downtime;
• incidents and near misses;
• overall workflow efficiency; and
• cost over the full life of the technology.

The Quality Dashboard Builder can help organisations structure this broader evidence rather than treating technology adoption as a standalone measure of success.

Scenario: organisation-wide data challenges an apparently successful rollout

A multi-site provider introduces similar service robots across several residential care homes.

Central reports show high overall usage and positive staff feedback.

When results are examined by location, one home uses the robots far less frequently than the others.

Management initially assumes staff resistance.

A local review produces a different explanation. The building layout contains narrow circulation areas and several points where the robots regularly obstruct movement during busy periods. Staff have gradually stopped using them because they interfere with residents moving through the home.

The problem is environmental rather than cultural.

The organisation does not impose a utilisation target. It redesigns routes, limits robotic deliveries during peak periods and reviews whether a different technology would be more appropriate for that building.

This demonstrates why governance needs variation data.

Low adoption may represent poor leadership or inadequate training, but it may also indicate that workers have identified a real limitation that central decision-makers have not seen.

Human relationships should be treated as a care resource

The most important limit on automation may ultimately be relational.

Care is not simply a bundle of tasks.

Conversation while helping somebody dress, reassurance during a transfer and encouragement during rehabilitation can all contribute to emotional security, confidence and wellbeing.

If robotics removes repetitive physical effort while preserving those interactions, technology can strengthen care.

If it removes the interaction itself, the consequences may be less positive.

This is especially important for people who have limited social contact outside formal services.

An apparently efficient automated visit may remove one of the few regular human interactions in the person’s week.

Productivity therefore needs a broader definition than minutes saved.

Robotics should not become a response to structurally inadequate staffing

Technology can support workforce capacity, but it should not legitimise unsafe workforce assumptions.

A care home cannot reasonably reduce staff below what people need and then rely on robots to compensate for absent human support.

Robots cannot take responsibility for an unexpected safeguarding disclosure, understand complex emotional distress or manage every emergency.

The stronger approach is to determine the human staffing model first and then identify where robotics can reduce burden or increase capability.

This keeps technology within the principle of workforce assurance rather than allowing it to become a substitute for it.

Funding models need to consider whole-life cost

Public funding has helped Hong Kong overcome the upfront cost of gerontechnology.

As adoption matures, greater attention will be needed on sustainability after acquisition.

Robotics may involve:

• maintenance contracts;
• software licences;
• connectivity;
• replacement batteries and components;
• staff training;
• cybersecurity updates; and
• eventual replacement or disposal.

A technology that appears affordable because acquisition is subsidised may create significant recurrent costs.

Service organisations therefore need whole-life financial analysis rather than relying on purchase price alone.

The developing gerontechnology market will increase provider choice

Hong Kong’s gerontechnology ecosystem is continuing to expand.

By late 2026, around 140 companies at Hong Kong Science and Technology Parks and Cyberport were reported to be involved in gerontechnology solutions and services.

The Government has also announced HK$100 million for a Gerontechnology Promotion Scheme intended to support local technology enterprises in developing and improving elderly-care products and solutions.

Greater supply should increase the range of options available to care services.

It also makes procurement competence more important.

Providers need to distinguish demonstrations from evidence, innovative features from necessary functions and short-term novelty from sustainable operational value.

Governance needs an explicit view of robotics

As robotics becomes more routine, leadership oversight should move beyond approving individual purchases.

Organisations should know where robots are used, what purpose they serve, what risks they create and whether intended benefits are being realised.

Material incidents, persistent faults, low adoption, staff concerns and service-user feedback should be visible through normal quality governance.

Leaders can use the Governance Maturity Assessment to examine whether accountability and assurance around significant technological change are sufficiently developed.

Robotics should not sit permanently within an “innovation” silo after deployment.

Once a robot becomes part of everyday service delivery, it becomes part of everyday operational governance.

International learning: adoption infrastructure matters as much as invention

Hong Kong’s experience offers a useful international lesson because its approach has not relied solely on technology developers.

Funding mechanisms, welfare organisations, residential services, government departments, community programmes and innovation institutions have all contributed to the gerontechnology ecosystem.

The institutional model is shaped by Hong Kong’s own welfare and innovation structures and cannot simply be reproduced elsewhere.

The transferable principle is that adoption needs infrastructure.

Inventing a robot is only the beginning. Care systems also need ways to test products, finance adoption, train workers, support implementation, evaluate outcomes and remove technologies that do not deliver expected value.

Other countries could adapt that principle without replicating Hong Kong’s specific funding mechanism.

The future will involve selective automation rather than robotic care systems

Robotic capability will continue to improve.

Embodied AI may allow machines to navigate complex environments, manipulate objects and interact more naturally. Rehabilitation robots may become more adaptive. Logistics automation may become increasingly routine.

Some of these developments are already emerging. Others remain experimental.

The likely future is therefore not a binary choice between human care and robotic care.

It is a continually changing division of work.

Tasks that currently require human effort may become automatable. That should create an opportunity to redesign human roles around judgement, communication, rehabilitation, reassurance and relationship.

The strategic risk is that services automate what is easiest to measure rather than what is least valuable for humans to do.

A thoughtful organisation will ask what people most need from another person and then use technology to protect capacity for those interactions.

Conclusion

Hong Kong has moved beyond treating gerontechnology as a small innovation experiment. Public investment, the Innovation and Technology Fund for Application in Elderly and Rehabilitation Care, technology-enabled residential settings and a growing local gerontechnology ecosystem have created increasingly practical routes for robotics to enter elderly and disability services.

The strongest applications are likely to be those that solve identifiable operational problems: reducing physically demanding work, improving internal logistics, extending rehabilitation, supporting independence and enabling staff to spend more time on activities requiring human judgement and relationship.

That does not make robotics automatically beneficial. Care organisations still need to address acceptance, safety, maintenance, workforce competence, data governance, building design and whole-life cost. Companion and interactive robots require particular care because their role reaches beyond efficiency into the person’s social and emotional environment.

The central strategic challenge is therefore not replacing human care. It is deciding which work technology can perform well enough that human capability can be used where it matters most. Hong Kong’s developing robotics landscape provides significant opportunities to redesign care around that principle. The measure of success should not be how visible robots become within services, but whether people gain greater independence, workers experience more sustainable roles and care becomes safer, more responsive and more human where human presence is irreplaceable.