Blog / AI Employees

AI Receptionist for Home Care: The 7pm Family Call

DependifyOctober 16, 20265 min read
AI Receptionist for Home Care: The 7pm Family Call

Home care enquiries do not arrive during office hours. They arrive at 7pm, after the adult daughter has finished work, visited her father, found the kitchen in a state, and decided that something has to change. She calls three agencies from the car. Two ring out. One answers.

That's the whole competitive dynamic, and it's why after-hours call handling is worth more in home care than in almost any other local service. The caller is at a decision point, they're calling several providers, and the first credible conversation usually wins the case.

What actually happens to your after-hours calls today

Most agencies have one of four setups.

Ring out. Some percentage never call back. There's no way to know how many, because a call that never connected leaves no record.

Voicemail. Better than nothing, worse than most owners assume. Families in the middle of a caregiving crisis often don't leave one, and those who do have usually called someone else before you hear it.

Forwarded to the owner's mobile. Works, at a cost you may have normalised. It's also inconsistent: the calls that arrive during dinner get handled differently from the ones that arrive at 9:30pm.

An answering service. Reliable pickup, generic handling. They take a message. They can't tell a family what your minimum hours are, whether you cover their zip code, or book a consultation — so the caller is still waiting until morning.

What an AI receptionist does differently

It answers every call, at any hour, and it can actually do things:

  • Qualify the enquiry. Hours needed, area, private pay or insurance, when care needs to start, what kind of help.
  • Answer routine questions. Minimum hours, coverage area, whether you take her father's situation, roughly what it costs — from your own answers, not invented.
  • Book the consultation into the calendar while she's still on the phone. This is the piece that separates it from an answering service: the family ends the call with an appointment rather than a promise.
  • Escalate properly. A caregiver reporting a fall, a family describing a medical emergency, an active client in distress — these route to your on-call phone immediately, not into a queue.
  • Handle Spanish, which in most of our target metros is not a nice-to-have.
  • Log everything, so the morning starts with a transcript and a booked appointment rather than three voicemails.

And it should say what it is. An assistant that identifies itself as an assistant converts fine. One that pretends to be a person creates a bad moment when the family works it out — usually in the consultation you just booked.

Where the line sits

We'll automate the front of the call. We won't automate the parts where being wrong costs someone something real.

Handled by the assistantAlways a human
Qualifying a new family enquiryAny clinical question about an active client
Answering published policy questionsA caregiver reporting an incident or injury
Booking a consultationTelling a family a shift is uncovered
Taking a detailed message with a callback timeA billing dispute
Routing an urgent call to on-callAny decision about care

The escalation rules should be strict rather than clever. If the caller mentions a fall, chest pain, confusion, a medication error, or anything the model isn't confident about, the correct behaviour is to hand off immediately — not to reason about it. A system tuned to be helpful in an emergency is a liability; one tuned to escalate fast is an asset.

What it's worth

Do the arithmetic on your own numbers rather than trusting a vendor's.

A private-pay client at $30 an hour and 25 hours a week is about $3,250 a month. Even at a conservative twelve-month length of service, that's roughly $39,000.

Now count the calls you currently miss. If you can't count them, that's the first finding — pull your carrier's call log for last month and look at after-hours calls with no answer and no callback. [Verify with your own data — this is the number that decides whether any of this is worth doing.]

If missing one enquiry a month costs an annualised $39,000, the case for answering the phone at 7pm doesn't need much salesmanship. And the same logic runs on the recruiting side: caregivers also call and apply in the evening, which is the other half of the after-hours problem.

Where this sits in the order of operations

Honestly: not first.

If your agency can't fill the shifts you already have, an AI receptionist books consultations for cases you'll have to decline — which damages the referral relationship more than a missed call would have. In that situation, staffing is the constraint, and the constraint is usually further upstream than owners expect: in the August 2026 Houston audit, five of five private-pay agencies had no JobPosting markup on their careers pages, meaning their open roles could not appear in Google's job panel at all. Check yours in fifteen seconds.

The sequence that works:

  1. Visibility — caregivers can find your roles
  2. Speed — you reach applicants before competitors do
  3. The phones — you stop losing families who called at 7pm
  4. Scheduling and the rest — call-out coverage, authorisations, follow-up

An AI receptionist deployed at step three is a revenue asset. Deployed at step one it's an expensive way to say no to more families. What each AI Employee owns, and what it's measured on, follows that same order.

Questions to ask any vendor, including us

  1. Does it identify itself as an assistant? If the answer is "it sounds completely human", ask what happens when the family finds out.
  2. What exactly triggers escalation, and can I see the list? Vague answers here are disqualifying.
  3. Does it book into my actual calendar, or take a message? Message-taking is an answering service at software prices.
  4. What does it say when it doesn't know? "I'll have someone call you back at 9am" is right. Improvising your pricing is not.
  5. Where do recordings and transcripts live, for how long, and who can access them? Family enquiries contain sensitive personal information even when they contain no medical record.
  6. What happens when it fails? Not if — when. There should be a fallback path to a human number.

Our answers are on the AI Employees page, and how we handle data and disclosure is written down on our compliance page rather than described on a call.

Related

Frequently Asked Questions

What can an AI receptionist do for a home care agency?

Answer every call at any hour, qualify the enquiry (hours needed, area, private pay or insurance, start date), answer published policy questions, book the consultation into your calendar, escalate urgent calls to the on-call phone, handle Spanish, and log a transcript so the morning starts with an appointment rather than voicemails.

Should an AI receptionist tell callers it isn't human?

Yes. An assistant that identifies itself converts perfectly well, and a family who realises they were deceived usually doesn't attend the consultation you just booked. In some contexts the disclosure is a legal requirement rather than a courtesy.

When should an agency not deploy an AI receptionist?

When it can't staff the cases it already has. Booking consultations for cases you'll decline damages referral relationships more than a missed call does. Fix caregiver visibility and applicant response speed first, then the phones.

The AI workforce

Hire one AI Employee. Add the next when the bottleneck moves.

One job function, run by AI, deployed into the operation you already have and accountable for one number — starting with the roles Google can't currently see.

See the AI workforce

Or run the free visibility check first — no signup.

Related to workforce

More on ai employees

Check my panel visibility — free