By Ravi Sundaar, Guest Author and Chief Technology Officer of Aretus Inc.

Adults 50+ control 70% of disposable income in the United States and represent the fastest-growing segment of new smartphone users. Yet as Agentic AI moves into consumer products, this demographic is at risk of being an afterthought, once again. The design decisions being made right now will determine whether that changes, while the technology is evolving.

AI usage among adults 50+ has grown from 18% to 30% in a single year [1], and Agentic AI is only beginning to reach mainstream consumers. For instance, about 70% of adults 50+ live with one or more chronic health conditions, yet only 13% currently use technology to help manage them [2]. According to AARP, trust, privacy, and data security are the top concerns for roughly half of older adults who hesitate to adopt AI tools. These concerns can still be incorporated into the design of agentic AI applications. The need is clear, however.

Touchscreens penalize older adults with limited fine motor control. App ecosystems are becoming complex and require navigating permissions, updates, and nested menus. Voice assistants like Alexa are reactive and narrow: despite their usefulness in some situations, they are inadequate for navigating a Medicare form or managing a medication schedule. AI agents have the potential to reduce that friction meaningfully. Natural language input, proactive task completion, and persistent context make them well-suited to seniors, of whom it is unfair to expect digital fluency. But, the design of AI agents needs to account for them.

An AI agent with access to health records, financial accounts, and personal communications is a high-value attack surface. The threat vectors are specific: prompt injection attacks that manipulate agent behavior, session hijacking via poorly secured APIs, and social engineering through voice interfaces where seniors are already the most targeted demographic for fraud. Bolting on security after launch is not a good strategy and may leave gaping holes.

Responsible Agentic AI architecture should require explicit user confirmation before the agent acts on sensitive data. Anomaly detection tuned to flag unusual requests should be a baseline feature, not an afterthought. Authentication needs to be senior friendly and low-friction. It may require the use of voice recognition or trusted-contact verification rather than password-based approaches. Privacy disclosures should be plain-language and surfaced in the interface, not buried in terms of service. These are well-understood design problems; the gap is prioritization.

The existing technology landscape has largely sidestepped this gap because the industry focuses on functions over security. Amazon’s Alexa has reach but limited agentic capability. Apple and Google are optimizing for the mainstream. Startups in the Agetech space have focused mainly on hardware and monitoring. Agentic AI for older adults remains largely unaddressed and the norms for how agents handle consent and fraud protection for this population are still being written, making this an unusually early moment to influence how it develops.

The decisions that shape how Agentic AI treats older users are being made now, in product reviews, design discussions, and research plans. The right questions to ask: Does the security model address the specific risks seniors face, or is it one-size-fits-all? Are privacy controls visible in the interface, or effectively hidden? And is the product being tested with users over 65, or is the team assuming younger cohorts generalize? These are easier to get right today than they will be in two years.

References

[1] AARP. AARP Tech Trends: Adults 50-Plus and Technology, 2025. AARP Research, 2025. https://www.aarp.org/research/topics/technology

[2] AARP. AARP Tech Trends Survey, 2024. AARP Research, 2024. https://www.aarp.org/research/topics/technology