The loneliness of old age is one of the least glamorous problems in modern healthcare, and one of the most stubborn. In aged-care facilities across the developed world, residents can go hours between meaningful conversations, while care staff — perpetually short-handed — triage their time toward the medical and the urgent. Into that gap steps a new and contentious category: embodied AI built explicitly for companionship.
Australia’s Andromeda, led by founder and CEO Grace Brown, is one of the startups betting that a robot can help. Its product, Abi, is designed to reduce loneliness among aged-care residents while supplementing, rather than replacing, the humans who do the caring. It is a deceptively simple pitch that opens onto some of the hardest questions in the future of work — about dignity, the limits of automation, and what we actually owe the people we care for.
What Abi Does
Abi is built around engagement. Rather than dispensing medication or lifting bodies — tasks fraught with liability and physical risk — the robot is oriented toward the social texture of a resident’s day: conversation, stimulation, prompts and reminders, and the small interactions that keep a mind active and a person feeling seen. In a sector where chronic loneliness is linked to cognitive decline and poorer health outcomes, that companionship layer is not a frill. It is arguably the part of care most often squeezed out when staff are stretched thin.
The deliberate framing matters. Andromeda positions Abi as a supplement to human carers, not a substitute. The robot is meant to absorb some of the engagement load — the chats, the check-ins, the reminders — so that human staff can spend their limited hours where their judgement and warmth are irreplaceable. It is labour augmentation, not labour replacement, and the company has been explicit about that distinction.
The strategy is now drawing serious capital. According to Startup Daily, which named Andromeda among five Australian startups to watch in 2026, the company raised roughly A$23 million in a round led by Forerunner Ventures, at a reported valuation of around A$100 million, to fund an expansion into the United States. That is a meaningful signal: investors with a track record in consumer and category-defining brands are treating companion care robots as a real market rather than a curiosity.
Why Care Is a Frontier for Embodied AI
There are good reasons embodied AI is converging on aged care, and they are mostly demographic. Populations across the developed world are ageing rapidly, and the workforce available to care for them is not growing to match. Aged-care providers everywhere report persistent staff shortages, high turnover, and burnout — structural problems that more recruitment alone has failed to fix. When demand for care rises faster than the supply of carers, technology stops being optional and starts being a release valve.
Care settings are also unusually well-suited to today’s robotics. The frontier of embodied AI is constrained by the messiness of the open world; robots still struggle in unpredictable, unstructured environments. An aged-care facility is the opposite of that. It is a controlled, mapped, relatively predictable space, with staff on hand to supervise and intervene. Residents follow routines. The physical layout is stable. That makes it one of the few real-world deployments where an embodied system can operate safely and usefully today, rather than in a demo reel.
This is why the augment-not-replace framing is more than PR positioning. It reflects an honest read of what the technology can and cannot do. A robot that tries to replace a carer will fail at the hardest, most human parts of the job and erode trust in the process. A robot that takes on the repeatable, social-engagement layer — under human supervision, in a structured setting — fits the technology’s actual competence. Andromeda is essentially designing to the grain of what’s possible, which is also the grain of what’s ethical.
The Hard Questions
None of this dissolves the genuine discomfort the category provokes — and it shouldn’t. The first and largest question is dignity. Offering a vulnerable, sometimes cognitively impaired person a machine for company can read as either compassion or abandonment, depending entirely on intent and execution. Is the robot there because it genuinely improves a resident’s day, or because it is cheaper than a human and lets a provider quietly cut hours? Consent complicates this further: residents with dementia may not fully understand what they are interacting with, raising real questions about who agrees to a robot’s presence and on what terms.
Then there is the question of where human care is simply irreplaceable. Some of what carers do can be supplemented. The deepest parts — being present with someone who is frightened, holding a hand at the end of life, the intuition that something is wrong before any sensor registers it — cannot. The risk is not that robots do this badly; it is that their presence creates an illusion of coverage that justifies pulling humans away from exactly these moments. The augment-not-replace promise only holds if providers honour it in their staffing decisions, not just their marketing.
That makes evidence the decisive issue. The claim at the heart of Abi — that it reduces loneliness and improves wellbeing — is testable, and it must be tested rigorously rather than asserted. Self-reported satisfaction, validated loneliness measures, engagement data, and longitudinal health outcomes should all be on the table, ideally scrutinised by independent researchers and not just the vendor. A companion robot that demonstrably lifts wellbeing is a public good worth funding. One that merely placates residents while papering over understaffing is a problem dressed as a solution. The category will live or die on which of those it turns out to be.
The India Read
For Indian founders, operators, and policymakers, the Andromeda story should land as a preview rather than a foreign curiosity. India is ageing — more slowly than the West today, but on a curve that bends sharply over the coming decades, producing tens of millions of older citizens. The traditional safety net of multi-generational households is also fraying as families urbanise, migrate for work, and shrink. The result is a quietly growing cohort of older Indians who are, in practical terms, alone for large parts of the day.
The catch is affordability and fit. A care robot priced for Australian or American facilities will not translate directly to a market where most eldercare is informal, family-provided, and cost-sensitive. The opportunity in India is unlikely to be the imported A$100-million-valuation hardware play; it is a cheaper, possibly software-led or lightly-embodied adaptation built for Indian homes rather than institutional facilities. Cultural fit matters just as much: companionship tech in India will need to handle multiple languages and dialects, religious and dietary routines, and family structures where the device augments relatives and domestic carers rather than salaried institutional staff.
The caution is the same one the West is wrestling with, only sharper. In a market where labour is comparatively cheap and human care abundant, the temptation to deploy robots as a cost-cutting substitute is weaker — but the risk of treating loneliness as a problem you can buy your way out of, instead of a social failure to address, is just as real. The smart Indian play is to treat care-tech as a tool that helps families and carers do better, watched closely for evidence, and never sold as a replacement for showing up. That is the line Andromeda is trying to walk in Australia. India will eventually have to walk it too — at far greater scale.
