Retention intelligence for home care agencies
Recognizing Unique and Brilliant You
RUBY sees the caregiver who is about to quit, and says the thing that keeps her.
Home care loses roughly three of every four caregivers each year, and every departure takes patient visits with it. No agency can hire its way out of that. RUBY watches the operational signals that come before a resignation, explains its reasoning to a human, and hands that person specific, dignified recognition drafted from the caregiver's own work.
The reality
Replacing a single caregiver costs about $2,600, which compounds to more than $423,000 a year for a 100-worker agency. Meanwhile the direct care workforce faces 9.7 million job openings between 2024 and 2034, and the median wage sits at $17.36 an hour.
The cruel part is that the caregivers agencies lose are often the ones who never wanted to go. In a survey of 849 home care workers, the answers were not about pay. They said helping people, and the one-on-one connection with my patient. Many described the work as a calling.
One aide with fifteen years at the same agency put it plainly: she was with a longtime patient the day he died, and nobody called to check on her. That is the moment RUBY exists to interrupt.
What RUBY does
Each week RUBY weighs call-off trend against a caregiver's own baseline, declining hours, silence since their last recognition, sustained high-acuity caseload, and tenure band, with recent shift pickups as a protective factor. It produces a 0 to 100 risk score and a band.
Every score comes with a plain-English reason and a confidence flag. Reach out to Loretta: hours down 45% and no recognition in three weeks, confidence high. Possible risk: new hire, only two weeks of data, confidence low.
RUBY scans recent shifts, last-minute pickups, hard-to-fill coverage, patient types, tenure milestones, and reliability streaks, then drafts recognition specific enough to be unfakeable. Sixteen years in, and you are still the one we count on.
RUBY monitors how recognition is distributed and flags any caregiver in the bottom decile relative to hours worked for weekly human review, so the quietly excellent stop being the quietly overlooked.
Opt-in check-ins tuned to each caregiver's archetype, pull whenever they want to talk, push only during shift hours, and always with a path to a real person. Peer tips turn mentoring moments into recognition for the caregiver who shared them.
Points and levels sit beneath the AI, with a required plain-language reason on every point. Caregivers see their own totals; peers see levels and badges only. Nothing earned is ever taken away, because a program that penalizes is a program that insults.
How it works
It connects to the payroll or EVV data an agency already keeps, and needs no new hardware from caregivers, who use the phone in their pocket.
The Recognition Champion is chosen for galvanizing energy rather than title. They work a ranked retention-risk queue and see why each name is on it.
Recognition goes out in one tap, as drafted or edited. Every AI output is logged for bias audit, and no message ever reaches a caregiver without a person behind it.
Who it serves
Direct care worker
Recognition and encouragement that names what they actually did, plus control over what RUBY shares about them and how it is allowed to reach out.
Recognition Champion
A ranked retention-risk queue with reasons and confidence, and ready-to-send recognition, so noticing stops depending on one overloaded supervisor's bandwidth.
Care organization
Retention, protected visit capacity, and an equity monitor over its own recognition culture.
Patients and families
A one-tap shout-out for a caregiver who made a difference, which becomes a Ruby Reward on that caregiver's profile.
In their words
“I'm ready to replace AVA right now with this. I don't want to wait years. Can we go faster?”
“I love having it all in one place in the palm of my hand. The only way I'm getting information is reading the on-call notes and you have to read 50 of them.”
“I don't get to see all of these reports. For me to have that in the palm of my hand about my girls — I love it. I love where it said they picked up how many shifts this week.”
“I think even just having like Ruby — just be like, man, I had a rough day. And just for someone to listen — I think it would help.”
“It shouldn't be like an obligation call. It should be like I care about you so I'm checking in. That's a big difference.”
What we are measuring
Where this comes from
AVA Rewards, co-developed by Laura Czuba, was deployed at a 60-location home care organization serving roughly 5,000 caregivers, and documented a 324% ROI and about $1.3 million in annualized benefit from reduced turnover and call-offs. What AVA could not do was notice on its own, which is exactly what RUBY's AI adds.
The 2017 survey of 849 caregivers produced five evidence-based archetypes that shape every design decision. A 2019 structured retrospective on the live program, and July 2026 design interviews and prototype testing with aides and supervisors, refined it further. Caregivers asked for a self-serve points balance, digital delivery, SMS-first contact, and the removal of a penalty they found undignified. All four are in the product.
RUBY is at TRL 3 with a running proof-of-concept of its two core AI components. Figures cited are drawn from Activated Insights 2025, PHI 2025, HHAeXchange 2025, Amesite 2024, KFF 2025, and Gallup and Workhuman 2024, alongside Hearth Care AI's own program data. Nothing here describes a shipped commercial product.
Tell us who you are and we will reach out when there is something real to try.