TL;DR

Every funeral SaaS vendor in 2026 mentions AI. Many vendors’ AI implementations are good and useful (Tukios obituary drafting, Vestamere’s case-record auto-population, Grace AI’s NFDA-award-winning workflow). Some are not (data harvesting from grief, AI-generated "digital avatars" of the deceased without family consent, automated upselling).

The funeral industry has two parallel conversations about AI:

• Vendors are pushing AI features hard because they’re a 2026 differentiator in sales.
• Practitioners (notably DNA Memorial’s Dr. Ryan Lehto in Funeral Technology and Trends) are pushing back: "Technology can bring comfort and connection during times of loss. But unchecked, it risks turning grief into just another revenue stream."

Both are right. This article: how to tell good AI from bad in funeral software, what to ask vendors before signing, and what Vestamere’s position is.

The 5 categories of AI in funeral software

1. Drafting assistance (LOW RISK)

AI drafts an obituary from intake data, drafts an arrangement summary, drafts a tribute video script. The human reviews and edits before anything goes out.

This is the safest category. The AI does the typing; the human owns the decision. Examples: Tukios obituary AI, Vestamere’s obituary engine (M3 with Cerebras / Anthropic / OpenAI / Groq providers; family + director review before approval).

What to ask the vendor:

• "Does the family or director review the AI output before it’s sent or published?"
• "If we don’t like the AI draft, can we override it without restriction?"
• "Is the AI output watermarked or fingerprinted in a way that affects how families see it?" (Hint: it should not be.)

2. Workflow automation (LOW RISK)

AI extracts structured data from unstructured intake (handwritten family notes → form fields), pre-fills standard paperwork, suggests case-management tasks based on case type. The AI is operating on a fixed input set.

Examples: Vestamere’s automatic family-intake field detection, Passare’s AI workflow suggestions.

What to ask:

• "Is the AI processing happening on-server (we control it) or is it sent to a third-party AI provider?"
• "If third-party (e.g., OpenAI, Anthropic), is family PHI being sent there?"
• "What’s your data-processing addendum (DPA) with that provider?"

3. Family-facing AI (MIXED RISK)

Chatbots that answer family questions, AI-generated condolence cards, AI-drafted grief support content delivered to the family.

This works when families opt in and the content is clearly AI-assisted. It fails when families don’t know they’re talking to AI, or when the AI is configured to upsell.

What to ask:

• "Are families told when content is AI-generated?"
• "Can families opt out of all AI interactions and talk to a human?"
• "Is the AI configured to mention products or services? If yes, how is that boundary controlled?"

4. Data extraction for monetization (HIGH RISK)

AI parses obituary text, social media of the deceased, family contact lists. The data is used for targeted advertising — sometimes the funeral home’s ads (less bad), sometimes third-party ads (very bad).

This is what Lehto warns about. Some "AI-powered obituary platforms" exist primarily to harvest data about the deceased and the surviving family for downstream marketing.

What to ask:

• "Is family data ever used for advertising, even your own?"
• "Is deceased data ever used for advertising?"
• "Is family contact information shared with third parties under any circumstance?"
• "What’s your data retention policy?"

If you don’t get clear NO answers to the first three, the vendor is in this category. Vestamere’s position: hard NO to all three.

5. AI-generated likeness of the deceased (VERY HIGH RISK)

"Digital avatars," AI-generated voices of the deceased speaking to the family, AI-generated photos showing the deceased in scenes they never inhabited. Some vendors offer this as a "memorial" feature.

Ethical considerations are severe. The deceased cannot consent. Surviving family may want it; future surviving family or other relatives may strongly object. The line between memorial and uncanny is family-specific.

What to ask:

• "Do you offer any features that generate likeness, voice, or content attributed to the deceased?"
• "If yes, what consent is required and how is it documented?"
• "Is the generated content stored or only ephemeral?"

Vestamere’s position: we don’t build this. We will partner with vendors who do this ethically (explicit family consent, clear "AI-generated" labeling, deletion on family request), but it’s not in our product roadmap.

What good vendor answers look like

A funeral SaaS vendor pitching AI in 2026 should be able to answer all of these without hedging:

1. Where does the inference happen? (Their servers? OpenAI? Anthropic? Open-source on-prem?)
2. What data is sent to the AI? (PHI? Family contact info? Or only de-identified data?)
3. Who owns the model outputs? (You, the funeral home. Not the vendor.)
4. Is the model trained on your data? (It should NOT be, unless you explicitly opt in.)
5. Can you turn AI off entirely? (Yes, if the product allows it. Some vendors require AI usage as part of their pricing.)
6. What happens to AI-generated content if a family complains? (Funeral home should be able to remove, edit, or delete entirely.)
7. Is there an audit log of AI usage? (Useful for FTC compliance and family transparency.)

If a vendor hedges on any of these, ask harder. If they can’t answer, walk away.

What Vestamere’s position is

We use AI in three places, openly:

1. Obituary drafting (M3 in our codebase). Cerebras, Anthropic, OpenAI, and Groq are configured as providers (free mode defaults to Cerebras for cost; paid mode lets the home choose). The AI drafts; the director and family review before approval. The AI output is marked as AI-drafted in the case record.

1. Family intake field detection. Unstructured family-provided text (life story brief, anecdotes, preferences) is structured into form fields. The director reviews the structured output. No PHI is sent to third-party AI without F2 personalization consent from the family.

1. Aftercare content personalization. Grief resource templates are personalized to the family’s situation (disposition type, relationship, declared religious orientation). This is template-driven with AI-assisted phrasing; never AI-generated likeness or content attributed to the deceased.

What we DON’T do:

• We don’t train our models on your data. Period.
• We don’t share family data with advertisers, ever.
• We don’t generate likeness, voice, or content attributed to the deceased.
• We don’t monetize aftercare touchpoints with third-party advertising.
• The deceased never receives any communication. Ever. (See voice-and-tone.md for the explicit principle.)

These aren’t marketing claims. They’re hard rules in our codebase. If we ever cross them, existing customers can leave with a full refund + their data export.

The vendor landscape in 2026 (honest take)

Vendors we’ve seen handle AI well:

• Tukios obit AI: discrete drafting tool, human-reviewed, no data harvesting evident.
• Grace AI (NFDA 2025 Innovation Award): workflow automation, FD-controlled.
• Impart (Melissa Schmidt — FD-led): FD-created and FD-aware ethics.

Vendors with mixed signals:

• Keeper Memorials AI obituary: real product, but the parent platform has obituary-data monetization features that Lehto critiques specifically.
• Some Passare AI features: vendor self-promotes hard; ask specifically about data-sharing terms.

Vendors to scrutinize hard:

• Any "AI-powered" obituary platform that doesn’t pay you (the funeral home) to use it. Free obituary platforms usually monetize the family data somehow.
• Any "digital avatar of your loved one" product, regardless of how well-intentioned.

The funeral industry trade press (Connecting Directors, Funeral Service Insider, Funeral Director Daily) routinely covers AI vendor launches. They’re generally positive — that’s their incentive. Practitioner critique (DNA Memorial / Lehto on LinkedIn) is where the ethics-aware perspective comes from. Both are worth reading before you sign anything.

The bottom line

AI in funeral software is here. Some of it is genuinely useful — saves directors hours, makes families’ experience better, frees humans for the actual humanity of the work. Some of it is genuinely harmful — turns grief into a data product, exploits families during their worst week, generates content the deceased never consented to.

The category isn’t good or bad. The vendor’s answers to the 7 questions above are good or bad.

Ask the questions before you sign. The funeral home that does this is the funeral home that doesn’t end up on a Lehto blog post in 2027.

Vestamere’s full AI architecture is documented at `src/lib/ai/` in the codebase. We’re happy to answer any of the 7 questions in writing — email founders@vestamere.com.

Sources: `docs/research/synthesis/director-pains.md` P11, `competitor-mentions.md` AI-only entrants section, and the LinkedIn discourse documented at `docs/research/raw/linkedin/technology-and-software.md`.