How Careflick combats senior loneliness through lifelike AI interactions with 54% user retention

How Careflick combats senior loneliness through lifelike AI interactions with 54% user retention

Learn how this innovative elder care platform partnered with Tavus to build AI humans that bring active engagement through emotional responsiveness and natural conversations.

“Many of the seniors we serve go four or five days without seeing another face, so our AI companion had to look and feel like a real person. Tavus provides the infrastructure that helps us make this possible.” – Abhishek Mohanty, Founder and CEO at Careflick

About

Careflick is an AI-powered platform that provides seniors with AI companions who help them age independently at home. Built in collaboration with caregivers, researchers, and clinical partners, the companions offer daily assistance, cognitive engagement, and emotional support.

Challenge

Building realistic AI companions to address senior loneliness at scale

Eight years in senior care provided Abhishek Mohanty with a critical insight: 86% of older adults want to live independently, but staying at home comes with long stretches of social isolation. When he founded Careflick to build AI companions that provide companionship and emotional support, he knew every AI agent had to feel human.

Careflick’s engineers had developed proprietary LLMs and cognitive analysis models to better understand seniors’ emotional and psychological needs. By analyzing real-time conversations, the system identified early signs of depression, anxiety, and cognitive decline, then crafted empathetic responses to seniors’ thoughts.

This orchestration layer ensured every interaction supported senior well-being. But insights alone weren't enough. To create meaningful connections, Abhishek needed a conversational video interface to bring the company’s AI to life.

In practice, Careflick’s AI agents had to see, listen, and respond with real perception—reading seniors’ tone and inflections, recognizing subtleties in facial expressions, and reacting to body language. With these capabilities, the AI can detect voice tremors and withdrawn expressions and respond accordingly—critical for seniors experiencing cognitive decline or emotional distress.

Latency was equally important. Any lags in the conversation break the sense of natural human interaction. And the visual experience needed to match. High-definition rendering that conveyed micro-expressions and eye contact would build trust with seniors who might be hesitant about technology.

To encourage daily engagement at scale, AI companions also had to adapt to seniors’ diverse realities. Some preferred speaking in their native languages, while others felt more comfortable talking to someone with a similar personality.

Because seniors prioritize convenience, all of these capabilities had to work seamlessly only on mobile apps. That way, seniors would access their companion with one tap, rather than navigating web interfaces that create friction.

Building a real-time video interface from scratch required specialized expertise, so tackling it in-house wasn't feasible. Instead, Abhishek began evaluating conversational AI video technologies. While some met the technical requirements, the avatars felt like text-to-speech animations, rather than true conversational companions with perception and emotional intelligence.

When a friend recommended Tavus, Abhishek tested one of their AI video agents directly on the website. The agent picked up on subtle visual cues like body language and gaze, then responded with natural timing and expression, creating the sensation of a real video call. Within minutes, he realized that the Tavus API provided the infrastructure for lifelike AI humans that seniors would want to spend time with.

“We were experimenting with a bunch of conversational AI products, but most of them felt scripted or robotic. Seniors need an experience they’d look forward to every day.”

Solution

Conversational Video Interface delivers natural, face-to-face AI experiences

Working closely with Careflick's engineers, Tavus integrated their modular Conversational Video Interface (CVI) with Careflick's LLM and TTS engines via developer-first APIs and SDKs. As a full behavioral stack, the CVI leverages three purpose-built models—Raven-1 for real-time perception, Sparrow-2 for conversational timing, and Phoenix-4.5 for emotional rendering—that deliver an end-to-end conversational experience.

With Tavus handling the conversational video layer, Abhishek secured the infrastructure to launch a tap-to-access mobile AI companion powered by Careflick's AI logic.

To better understand how seniors would engage, Careflick initially rolled out both voice-only interactions and Tavus-powered video agents. The results were decisive: video won 97% of the time, proving that genuine connection demands visual presence. With this clarity, Careflick removed the audio-only option and made Tavus’ face-to-face AI humans the primary interface.

For every interaction, Tavus’ Raven-1 reads seniors’ facial expressions, tone, and visual cues, building a multimodal picture of their emotional state and intent. Using this context, Careflick's AI logic shapes the response based on cognitive analysis, and lets Sparrow-1 and Phoenix-4 take over delivery. Sparrow-1 adapts to the speaker’s tone and mood, then determines when to speak based on the conversation’s natural rhythm. When the moment arrives, Phoenix-4 renders the companion’s expression with appropriate emotion and active listening signals.

When seniors open the app, this setup creates an experience that feels like a video call with a friend. As users share about their grandchild’s visit or health concerns, the AI human engages in thoughtful, lifelike dialogue, maintaining perfectly synced lip movements and natural vocal inflections.

These conversations flow seamlessly. Sparrow-1 adapts to different speech patterns in real time, distinguishing thoughtful pauses from finished thoughts and adjusting to mid-sentence redirects. Phoenix-4 matches this awareness visually, rendering full-head video at 1080p and 40fps with natural micro-expressions drawn from thousands of hours of real conversation data.

The CVI’s modular architecture gives Careflick granular control over every aspect of the AI companions. With this flexibility, Abhishek and his team create personas with distinct accents, languages, and personalities. For example, one companion delivers quirky comebacks in a Southern accent, while another opens every conversation with a signature catchphrase. 

This level of customization enables every user to find a companion they connect with, driving sustained engagement. In turn, this success enables Abhishek and his team to expand their reach.

The partnership extends well beyond implementation. If Careflick’s engineers have a question, they simply Slack the Tavus team and receive support within a day. In Abhishek’s words, Tavus has become a true extension of his team, enabling them to quickly iterate on new features and address feedback, keeping the focus on improving the senior experience.

"Tavus enables us to create AI companions that seniors genuinely look forward to talking to. We retain more users over time, which helps us combat loneliness at scale.”

Results

Careflick users spend 24 minutes daily with AI companions powered by Tavus

With Tavus, Abhishek and his team deliver realistic AI companions that engage seniors through perception, empathy, and emotional intelligence. This approach transforms how seniors experience aging at home, fueling growth that will double their user base.

As Careflick scales and enters new markets like India, Tavus’ video rendering and perception capabilities will be central to creating more meaningful interactions that support independent aging.

“It feels like Tavus is a part of Careflick—our teams are in absolute sync. They do everything to ensure we are supported and advancing on our mission.”