The Five Objections to AI Answering a Funeral Home's Phone

Earlier this year, the most established answering service in funeral care published a thoughtful piece on AI in funeral home answering. Their conclusion: “there really is no place for AI in funeral home answering.” The piece lays out five objections — brand damage, lack of emotional intelligence, failure to adapt mid-call, risk of misunderstandings, and ethics.

I build an AI answering layer for funeral homes, so you would expect me to argue with that piece. I’m not going to. Every one of those five objections describes a real failure mode, and I have heard versions of all five from funeral directors I respect. What I want to do instead is take each objection seriously and show what I built for it — including the two places where my answer is “you’re right, and that’s why a human takes over.”

1. “One robotic first call can send a family to another funeral home”

This is true, and it cuts both ways. The first call sets the tone for everything that follows, and a family that hears something cold or canned at the worst hour of their life will quietly call the next home in the search results. The same is true of a voicemail greeting.

So the bar is not “good for a machine.” The bar is your funeral home’s own voice. Here is how I hold it:

• The voice is trained on your home’s tone, and it does not take a single family call until you have personally heard and approved every greeting, every escalation, every core conversation.
• It never says the same sentence twice. Word-for-word repetition is the surest tell of a machine, so the voice is built to phrase things a little differently every time, the way a person does.
• Every call produces a transcript and a recording in your inbox within sixty seconds. There is no black box. If something ever sounds off, you hear it before the next family does.

And because trust is earned on your calls, not mine: I’d rather you hold it to that bar yourself — hear it before you commit to anything.

2. “A machine cannot genuinely feel empathy”

Correct. It cannot, and nothing I build pretends otherwise.

What a well-built voice can do is behave the way the best first-call training teaches: condolence first, then an actual pause. One short sentence at a time. Silence left alone, because a pause is a caller finding words, not a gap to fill. If the family starts telling the story of what happened, the voice lets them — the questions can wait, being heard cannot. The person who died is called by their name, gently, in the past tense. Never “the body.”

But the deeper answer to this objection is structural, not behavioural. When a call needs real comfort — raw grief, a crisis, anything beyond gathering details with care — the job of the voice is to bring in your on-call director, not to perform empathy it does not have. The design assumes a machine should not carry the hardest human moments. Its job is to make sure the person who can carry them is reached, quickly, with nothing lost on the way.

3. “Scripts can’t adapt when a call goes sideways”

The failure being described here is real, and anyone who has been trapped by a phone tree knows it. Decision-tree bots fall apart the moment a caller does not follow the script.

This is not that. There is no script to fall off. The voice follows the caller’s lead: a death report, a pre-planning question, a vendor confirming a delivery, a wrong number — each is handled as what it is. It gently establishes who is calling and adjusts: a family member is walked through the first-call details; a neighbour calling on a family’s behalf is treated kindly but nothing is promised that requires the family’s authority. If it sounds like a death may have just happened and no one medical has been involved, the voice stops everything and points to emergency services first.

And the honest backstop, again: anything truly off-pattern goes to your director, with a complete handover of everything gathered so the family never repeats themselves.

4. “Speech recognition fails on accents and distressed voices”

This is the strongest objection on the list, and it is exactly why I build multilingual first.

The common failure is not really about technology — it is a system forcing a distressed caller to operate in English. A daughter who has just lost her father does not want to navigate an English-only line in her second language at 3am. The answer is not better tolerance of accents. The answer is answering in her language: English, French, Punjabi, Hindi, Mandarin, Urdu, and more. The accent problem largely disappears when the call happens in the language the family actually speaks — something no answering service in this market offers beyond Spanish, at any price.

For what remains, accuracy is engineered, not assumed: every phone number is spoken back as words, digit by digit, and confirmed. Every name is spelled and confirmed. The director gets the recording, so nothing depends on a machine’s interpretation — you hear the family’s own words.

5. “Letting a machine handle a death call feels wrong”

This one is about values, so the only honest response is to state mine and hold them.

The voice never claims to be a person. If a caller asks whether they are speaking to a human, it discloses honestly. It tells callers the line is recorded, up front, on the call. It never makes arrangements, never advises a family, and never quotes a final price — that work belongs to a licensed funeral director and stays there. It is an intake layer that stands between a grieving family and voicemail, and it makes sure your licensed director is reached right away.

Because that is the real comparison. The choice at 3am is not a machine versus a compassionate human. Where a trained human answers every overnight call in the family’s language, that home should keep them. The choice, for most homes, is a machine that answers with care versus a voicemail box that answers with a beep — and the piece I linked above concedes this exact case, noting that for homes relying on voicemail after hours, an automated answer is a practical step up.

That is the whole position. I did not build this to replace the people in funeral service who answer with compassion. I built it so no grieving family reaches voicemail, in whatever language they speak.

Hear it instead of taking my word

Every claim above is testable in two minutes. The live demo answers as a sample funeral home — or call it directly at (647) 250-0214, from any phone. Speak to it the way a family would. Report a loss. Interrupt it. Switch languages mid-call. Ask it if it’s human.

And if you run a funeral home, the fairest test is the one I offer every prospect: hear it on a real call, against whatever answers your phone today. If the objections above show up when you test it, you will hear them — and you should walk away.