The AI Caregiver Economy

The 2025 AARP and National Alliance for Caregiving report puts the number of family caregivers in the United States at 63 million, a 45 percent increase from the 2015 figure of roughly 43 million. One in four American adults is now providing unpaid care to a family member or friend. Of those 63 million, 59 million are caring for an adult, and 44 percent report providing high-intensity care involving complex medical tasks such as managing infusion equipment, administering injections, or operating respiratory devices. Only 22 percent of those performing clinical tasks report receiving any formal training to do them. Nearly one in five caregivers reports fair or poor health attributable directly to the caregiving role. Half have experienced a major financial impact: depleted savings, accumulated debt, or inability to afford basic needs.

These are the people on whose labor the aging-in-place policy agenda actually rests. Every Medicare model that substitutes home-based management for institutional care, every FIDE SNP that coordinates LTSS in the community, every AHEAD hospital that avoids an admission by ensuring a patient has adequate support at home, depends on an informal caregiver being present, functional, and capable. Medicare does not pay them. It does not train them. It does not screen them for burnout. It does not track them as a policy variable. They are the invisible infrastructure beneath every model that assumes home as the default site of care.

What Medicare Covers and What It Does Not

#

Medicare’s caregiver coverage gap is structural, not incidental. The program was designed in 1965 around the concept of a Medicare beneficiary who receives professional medical services. The caregiver who makes those services possible by managing the beneficiary’s medications, coordinating appointments, managing behavioral responses in dementia patients, and performing nursing-adjacent tasks at home is not a covered person under the program. There is no Medicare benefit category for caregiver training, caregiver respite, or caregiver health assessment.

Medicaid fills some of this gap for the dual eligible population. Consumer-directed attendant programs in many states allow beneficiaries to hire and direct their own personal care workers, including in some states family members who become paid caregivers through Medicaid. Eleven million of the 63 million caregivers counted in the 2025 AARP-NAC report receive some form of compensation through Medicaid, VA, or other state programs. These paid caregivers are disproportionately younger, lower-income, and from communities of color. The policy implication is that the caregiving workforce that Medicaid compensates looks nothing like the caregiver population that Medicare policy assumes is available to support beneficiaries at home.

The HRSN framework that the Biden administration built between 2021 and 2024 created a Medicaid waiver pathway for states to fund social supports that address the conditions enabling community-based care: housing stability, nutrition, home modifications, and transportation. Eighteen states obtained approved Section 1115 waivers under that framework. CMS rescinded the governing informational bulletins and the HRSN Framework on March 4, 2025, reverting to case-by-case review of new waiver applications under the Social Security Act without the structured guidance. Existing approvals were not nullified, but states with pending applications lost the regulatory roadmap that had made HRSN waiver design tractable. The policy infrastructure for funding the social supports that make caregiving sustainable was narrowed at the same moment that the caregiver population was growing at its fastest recorded rate.

MAHA ELEVATE and the Social Connection Pillar

#

MAHA ELEVATE, which begins September 2026, includes social connection among its six health pillars alongside nutrition, physical activity, sleep, mental health, and substance use. The model funds up to 30 cooperative agreements across three years with approximately $100 million total, targeting chronic disease prevention and health promotion interventions for Original Medicare beneficiaries that are not currently covered under the program. The Notice of Funding Opportunity was expected in early 2026.

The social connection pillar reflects a policy acknowledgment, belated but real, that isolation and loneliness are clinical risk factors with measurable mortality and morbidity consequences. The Surgeon General’s 2023 advisory on the loneliness epidemic cited evidence linking social isolation to increased risk of heart disease, stroke, dementia, depression, and premature death at effect sizes comparable to smoking fifteen cigarettes per day. For Medicare beneficiaries, who are disproportionately represented among isolated populations due to mobility limitations, geographic distance from family, spousal bereavement, and age-related sensory impairment, social connection is not a wellness amenity. It is a determinant of clinical outcomes that the fee-for-service payment system has never had a mechanism to address.

MAHA ELEVATE does not create a permanent Medicare benefit for social connection programming. It creates a funded pilot that will gather cost and quality data to inform future coverage decisions. The distinction matters. A three-year pilot reaching 30 organizations is not a delivery system for social connection at Medicare population scale. It is a data-collection exercise that may, if the evidence is strong and the political conditions favor it, eventually support a coverage determination. The beneficiaries who need social connection support in the years before that determination are not served by the existence of the pilot.

AI Tools in the Caregiver Context

#

The AI product categories that are actually relevant to caregiving do not map cleanly onto the categories that Medicare policy has begun to regulate. CMS’s February 2024 FAQ memo on AI in prior authorization established that AI cannot serve as the sole basis for a coverage denial and that individual clinical circumstances must be considered. The proposed CY2026 rule’s AI guardrail provisions, which the Trump administration dropped from the April 2025 final rule, would have defined automated systems and patient care decision support tools in ways that created regulatory structure for AI in plan decision-making. Neither of these policy instruments addresses AI in the home as a caregiving support.

The relevant AI categories for the informal caregiver economy are care coordination platforms, safety monitoring systems, and caregiver education and task support tools. Care coordination platforms address the scheduling, medication management, appointment tracking, and cross-provider communication functions that caregivers of high-complexity patients perform manually and often poorly. The research on caregiver medication errors is consistent: caregivers managing multiple medications for a patient with multiple chronic conditions make errors at rates that produce clinically significant harm, and the errors are concentrated in dose timing, drug interaction recognition, and PRN administration decisions. An AI tool that flags interaction risks, sends adherence reminders calibrated to the patient’s actual schedule, and surfaces instructions in plain language reduces a real harm that Medicare currently does not measure or address.

Safety monitoring systems for the home environment include fall detection, wandering detection for dementia patients, and environmental sensors that flag anomalies in daily activity patterns. These technologies do not require any Medicare coverage to be deployed, and many are already in use. Their policy relevance lies in the question of whether their outputs can be integrated into Medicare-covered care management workflows. A fall detection alert that reaches a remote monitoring platform connected to an ACO care manager, who can then dispatch a home health nurse to assess injury risk and medication effects, is functionally different from a fall detection alert that rings a caregiver’s phone. The former requires interoperability between consumer-grade monitoring infrastructure and clinical workflow systems. The latter requires only a smartphone. Medicare’s 2026 RPM code set and the CCM billing framework create partial pathways for the former, but they require a billing provider and a clinical workflow that most informal caregiver situations do not have.

Caregiver education tools are the least regulated and most immediately deployable AI application in this space. A caregiver managing a parent with Parkinson’s disease who uses an AI platform to understand symptom progression, medication side effects, fall prevention techniques, and available community resources is not performing a Medicare-billable activity. But the platform may reduce avoidable hospitalizations, emergency department visits, and the escalating intensity of caregiver burden that eventually produces caregiver breakdown and beneficiary institutionalization. None of this generates a Medicare billing event that would sustain the platform commercially within the traditional fee-for-service payment structure.

The Commercial Model Problem

#

The absence of a Medicare payment mechanism for AI caregiver support tools creates a commercial model problem that policy is not addressing. The populations who most need caregiving support tools are lower-income, have limited digital literacy, and are caring for beneficiaries with high medical complexity. These are not the populations that consumer wellness markets serve profitably. The companies that can sustain investment in caregiver AI at the necessary scale and depth are those that can either find a payer relationship that generates recurring revenue or serve a population with sufficient income to pay out of pocket.

The payer relationship pathway runs through Medicare Advantage supplemental benefits and FIDE SNP value-based contracts. MA plans have used supplemental benefit flexibility to fund caregiver support as a strategy for managing member cost of care, and some plans have contracted with technology vendors to deliver caregiver navigation, coordination, and education services. The supplemental benefit contraction in 2025 and 2026, driven by CMS tightening its oversight of how plans price and administer VBID benefits, reduced the funding available for these programs. FIDE SNP contracts, where the plan bears full Medicare and Medicaid risk, provide the strongest financial incentive to invest in caregiver support because the full cost of caregiver breakdown, including emergency department use, inpatient admission, and institutional long-term care placement, flows through the plan’s capitated payment.

ACOs operating under MSSP and ACO REACH have care management infrastructure that can incorporate caregiver support into high-risk patient management programs, funded by the shared savings those programs generate. The AHEAD model’s global budget creates the same incentive structure at the hospital and health system level. In these accountable care environments, the economic case for caregiver support investment is documentable: a caregiver who is better trained, better coordinated, and less burned out produces fewer avoidable utilization events for a beneficiary population whose total cost falls within a shared savings or global budget accountability structure.

What Policy Can Enable

#

The policy mechanisms that would most directly expand the AI caregiver economy are not primarily AI-specific. They are payment and coverage mechanisms that create billing pathways for caregiver-adjacent services that already exist without adequate funding.

A Medicare caregiver training benefit has been proposed in multiple legislative vehicles and has not passed. Such a benefit would create a billable event for structured training of informal caregivers in clinical tasks, generating the revenue stream that would sustain both the training infrastructure and the digital tools that support ongoing caregiver competency. The RAISE Family Caregivers Act, signed in 2018, required HHS to develop a national strategy to support family caregivers. The strategy was released in 2022. Its implementation has not produced a Medicare benefit, a payment code, or a funded program of meaningful scale.

The Credit for Caring Act, which AARP and NAC have actively backed, would create a federal tax credit of up to $5,000 for working family caregivers covering a portion of their out-of-pocket caregiving expenses. A tax credit does not create a Medicare payment pathway, but it reduces the financial strain that drives premature caregiver exit from the workforce, which is the structural condition that produces beneficiary institutionalization when informal care infrastructure collapses.

Within existing Medicare authority, the CCM and RPM billing frameworks create the most tractable near-term pathway for integrating caregiver support into covered services. Chronic care management billing under CPT 99490 and the principal care management codes requires 20 minutes of clinical staff time per month for a patient with two or more chronic conditions. The clinical staff time can include coordination with informal caregivers as part of the care plan. RPM data that flows from a patient’s home environment can trigger clinical escalations that include caregiver notification and instruction. These pathways do not pay for the caregiver’s time or the AI platform the caregiver uses, but they create a clinical workflow context in which caregiver-facing technology can operate adjacent to covered services.

The most direct policy enabler for AI in the caregiver economy is ACO and global budget penetration. Every percentage of the Medicare population attributed to an accountable care entity with genuine shared savings or global budget risk is a percentage of the population whose care managers have a financial reason to invest in caregiver support infrastructure. AHEAD’s expansion to additional states, MSSP’s continued growth, and ACO REACH’s population health model create the financial environment in which caregiver AI can find a payer relationship. The technology is not waiting on policy. The commercial model is.

Related Reading

#

MCR-01_06 MAHA ELEVATE: Lifestyle Medicine Enters the Medicare Payment Lexicon MCR-07_05 Staying Home Longer: How Medicare Policy Is (and Isn’t) Supporting Aging in Place