Voice AI for Elder Care: A Proof of Concept with PacSana
Client
PacSana
Duration
Proof of Concept
Service
AI Automation - Voice AI
Key Results
7 Days
From brief to a working, testable proof of concept validating the approach at speed without a lengthy build cycle
3 roles
Role-aware responses built for carers, family members, and seniors in one system, personalised to each user type
GDPR + HIPAA
Compliant architecture designed from day one and ready for future scaling without a costly privacy retrofit
Validated
Voice AI confirmed as a viable, high-quality interaction channel giving PacSana clear evidence to guide what to build next
Where they were
when we started.
PacSana's team had a clear hypothesis: voice AI could make their care data far more accessible, especially for older adults, carers, and families who find dashboards and apps challenging. But before committing to a full product build, they needed to know whether current AI tools could actually deliver an experience their users would trust. The questions were practical: could it handle open-ended, emotionally-loaded queries? Could it be built to GDPR and HIPAA compliance standards from day one? And could it genuinely serve such a diverse set of users, tech-wary seniors, busy carers, anxious families, and business owners all at once, without feeling clinical or cold?
What we did
and why it worked.
- 1
Designed a secure, voice-first interface using Vapi (Whisper STT + natural-sounding TTS) to ensure the experience felt genuinely conversational and not robotic or transactional
- 2
Integrated GPT-4 for flexible conversational logic, able to interpret open-ended health-related queries with nuance and respond naturally in real time
- 3
Built custom middleware to simulate queries against PacSana's real-time mobility data, keeping the proof of concept grounded in realistic data flows rather than mocked outputs
- 4
Implemented role-aware responses tailored to different user types such as carers, family members, and the individual senior, so each person received relevant, appropriately framed information
- 5
Designed with privacy first from day one: GDPR and HIPAA-compliant architecture built into the foundation, not retrofitted at the end
- 6
Tested against real-world scenarios like daily movement summaries, sleep quality queries, fall prevention advice, and product questions and all delivered through natural, unhurried conversation
Key Scenarios Tested
"I need to understand how my mum has been today"
"How was her movement yesterday and sleep last night?"
"Can you give me more information on PacSana and the bracelet?"
"What can I do to avoid a fall?"
What changed
after the work.
7 Days
From brief to a working, testable proof of concept validating the approach at speed without a lengthy build cycle
3 roles
Role-aware responses built for carers, family members, and seniors in one system, personalised to each user type
GDPR + HIPAA
Compliant architecture designed from day one and ready for future scaling without a costly privacy retrofit
Validated
Voice AI confirmed as a viable, high-quality interaction channel giving PacSana clear evidence to guide what to build next
The findings that
shaped what came next.
- →
Voice interaction is viable with current AI tools - even for older adults - if designed with empathy and clarity
- →
GPT-4 can interpret diverse health-related queries with surprising nuance, but benefits from structured guidance and fallbacks
- →
System architecture matters: the ability to segment user roles and handle private data securely is critical for future scaling
- →
Experience quality is high - natural voice responses created a sense of comfort and trust, which is key for adoption in elder care
What this meant
for the business.
- →
Validated that voice AI can support future product directions
- →
Informed the roadmap with realistic capabilities and constraints
- →
Identified gaps in infrastructure to be addressed before full rollout
- →
Provided internal stakeholders with a tangible demo to guide planning and investment
The principles behind
the result.
- →
Tight focus on real-world tasks rather than broad, speculative use cases
- →
Ethical, privacy-aware design from day one - not bolted on at the end
- →
Use of best-in-class AI tools, carefully orchestrated to fit the care context
- →
Spoke directly to each user's real-world needs: busy carers, anxious families, tech-wary seniors
- →
The assistant was designed with empathy, clarity, and reliability - not just intelligence
What this could mean
going forward.
- →
The ability for PacSana to differentiate with a voice-first solution in a competitive market
- →
Improved customer retention and satisfaction by offering human-like intelligence at scale
- →
Reduced friction for elderly users, allowing them to access data using the most natural technology they already know - their voice
The bottom line.
This proof of concept provided PacSana with clear evidence of what voice AI can offer - not just as a novelty, but as a potential pillar of user interaction. It clarified both the opportunity and the readiness of the technology, allowing the company to make informed decisions about what to build next. Voice AI isn't just possible in elder care - it's powerful, when done right.
“It was great working with Scott. I briefly outlined our business needs and he ran with it, turning around a very impressive proof-of-concept in just a few days. It was exactly what I needed, allowing me to quickly validate the concept with our customers and investors.”