Funeral Home AI Policy

A funeral home AI policy should tell staff which tools they may use, what information they may enter, what work needs review, and which decisions remain with an authorized person. Begin with a small set of approved tasks. A policy that names real examples is easier to follow than a general instruction to use artificial intelligence responsibly.

This guide offers an internal policy starting point for funeral directors and managers in Illinois, Missouri, Kansas, and elsewhere. Generative AI produces content such as draft text from a prompt. The operating boundaries below are recommendations, not a claim that one national rule governs every tool or use.

Start the funeral home AI policy with approved tasks

List a few tasks staff may perform with approved tools, such as drafting a generic meeting checklist, simplifying an already public explanation, or practicing a response with fictional details. Name the person who approves additional uses. Separate preparation of a draft from permission to send, publish, or act on it.

For example, staff could ask for clearer wording for a blank appointment reminder, then have a colleague check it. Entering a real family's correspondence to generate a reply is a different use and needs a separate information-handling decision. Do not let a tool's convenience quietly expand the original approval.

Decide what information may enter each tool

Review the tool's actual terms, account controls, retention, access, and data-use settings before allowing staff to use it with nonpublic information. Identify who owns the account and can change those settings. A paid subscription or an “enterprise” label should not substitute for understanding the specific arrangement the business has accepted.

As a conservative starting policy, keep family records, payment information, identification documents, staff records, and private conversations out of unapproved tools. Use invented examples in training. The FTC's Start with Security guide describes avoidable risks from using real personal information in training and development when fictitious information could serve the purpose.

Require a reviewer to check the actual source

Assign a person to verify names, dates, service details, prices, and any legal or factual statement before a draft is used. Check against the authorized case record or a current primary source, as appropriate. A fluent sentence is not evidence. The reviewer should be able to explain where a consequential detail came from.

NIST's Generative AI Profile describes confabulation: confident output that is false or erroneous. It also recommends reviewing and verifying generated sources and citations. A link that looks plausible should be opened and checked for the precise claim; asking the same tool whether it is correct is not an independent check.

For a family-facing draft, review tone as well as facts. Remove invented memories, assumptions about beliefs, and feelings the family has not expressed. Keep any family approval already required by the normal workflow. Using AI to assist with wording should not bypass that approval.

Keep authority and public disclosure clear

State which uses are outside the approved scope. A practical starting policy excludes autonomous authorization of disposition, decisions about disputed instructions, individualized legal or medical advice, and automatic sending of case-specific messages. Tools can help prepare information for review; the person with the relevant responsibility still makes and records the decision.

If the firm introduces an AI interaction for visitors, explain that it is AI, describe its limits accurately, and provide a usable route to a person. Do not invent professional credentials or present a generated avatar as the director. Review the actual visitor experience and applicable requirements before launching it; an internal drafting policy does not approve a public chatbot.

Practice a mistake and make reporting simple

Use a fictional exercise containing one wrong date, one unsupported claim, and one detail that should not be entered into an unapproved tool. Ask staff to identify each issue and explain the next action. The exercise should test judgment about the output and the information, not simply whether staff can write a prompt.

Name the person who handles accidental uploads or incorrect published output. Staff should know how to report the incident promptly and preserve the facts without recirculating private information. Review approved uses when a tool, account setting, or workflow changes. Expand the policy only when the team can maintain the same standard of review.